Prior Authorization and Association With Delayed or Discontinued Prescription Fills

Abstract

PURPOSE

Prior authorization requirements are increasing but little is known about their effects on access to care. We examined the association of a new prior authorization policy with delayed or discontinued prescription fills for oral anticancer drugs among Medicare Part D beneficiaries.

METHODS

Using Medicare part D claims data from 2010 to 2020, we studied beneficiaries regularly filling one of 11 oral anticancer drugs, defined as three 30-day fills in 120 days preceding the plan's prior authorization policy change on that drug and continuously enrolled in the same plan for 120 days before and after the policy change at the start of a new year. The control group consisted of beneficiaries meeting the same utilization criteria, but who were enrolled in plans at the same time that did not implement a prior authorization policy change. The outcomes of interest were discontinuation of the drug within 120 days (analyzed with regression analyses) and time (in days) to next fill after a prior authorization policy change (analyzed using a quasi-experimental difference-in-differences event study).

RESULTS

The introduction of a new prior authorization on an established drug increased the odds of discontinuation within 120 days (adjusted odds ratio, 7.1 [95% CI, 6.0 to 8.5]; P < .001) and increased time to next fill by 9.7 days (95% CI, 8.2 to 11.2; P < .001), relative to patients whose plans did not have a prior authorization policy change.

CONCLUSION

Introduction of a new prior authorization policy on an established drug regimen is associated with increased probability of discontinued and delayed care. For some conditions, this may represent a clinically consequential barrier to access. Waiving prior authorization for patients already established on a drug may improve adherence.

INTRODUCTION

Prior authorizations are a common form of utilization management wherein a proposed treatment must be submitted to a payer for review before coverage is approved. The goal of prior authorization is to verify a patient's medical need for a given treatment. Use of prior authorization has increased over the past decade,^1-3 and clinicians, patients, and regulators have identified it as a common frustration and at times a barrier to access.^4-7 While financial barriers to accessing prescription medications are well-documented,^8-11 there is less evidence examining the impact of nonfinancial barriers such as prior authorization.

CONTEXT

• Key Objective

• Do nonfinancial costs—such as prior authorization—affect timely access to care?

• Knowledge Generated

• For Medicare Part D beneficiaries consistently filling an oral anticancer drug, the introduction of a new prior authorization policy on that drug increased the odds of discontinuation within 120 days and was associated with an average delay of 10 days in refilling the first prescription after the policy change, compared with Medicare beneficiaries whose plans did not change the prior authorization policy on those drugs. The clinical implications of discontinuations or delayed fills likely vary by drug.

• Relevance (S.B. Wheeler)

• This important analysis associates oral anticancer drug discontinuation and delayed prescription fills among Medicare enrollees with prior authorization requirements. Such prior authorization requirements by insurers may introduce unnecessary administrative burdens that reduce timely access to appropriate treatment and should be carefully addressed through policy change considerations.*

Although prior authorization can be burdensome,¹² rising prices pose challenges to individual patients and health care spending overall.^13-18 In some cases, utilization management tools such as prior authorization may help patients avoid unnecessary care and expense.¹⁹ An analysis estimating the spending effects of prior authorization in Medicare Part D found that spending averted substantially exceeded prior authorization's administrative costs.²⁰

Studies of prior authorization for medical services and physician-administered drugs find that oncology is among the most burdened clinical areas.^7,21,22 Orally administered cancer drugs have transformed cancer care in recent decades; they are also famously expensive.²³ Oncology drugs are one of five protected classes in Medicare Part D, meaning formularies are required to cover all approved drugs in that category. In this context, coverage restrictions, such as prior authorization, are the principal (only) strategy available to health plans to constrain utilization.

Adherence to oral anticancer drugs is important. These drugs are taken daily akin to other chronic conditions. Unplanned gaps in adherence are concerning as they can contribute to disease progression or drug resistance, which can exhaust available treatment options.^24,25 By design, prior authorization introduces administrative burdens on clinicians and patients to monitor or modify utilization, but it may lead to delayed or foregone care. This study uses Medicare Part D claims data from 2010-2020 to examine the consequences of introducing a new prior authorization policy on delayed fills or discontinuation of oral anticancer drugs. Medicare covers 18.4% of the US population, with beneficiaries consisting primarily of adults older than 65 years. The median age of people diagnosed with cancer is 66 years, giving Medicare a prominent role in cancer care coverage policy.^26,27

METHODS

Study Design, Data, and Study Population

We used regression and a difference-in-differences event study design to estimate the association of a new prior authorization policy with delayed and discontinued prescription fills among Medicare Part D beneficiaries for 11 oral anticancer drugs with at least 10 observed instances of new prior authorization requirements. The difference-in-differences design is well suited for examining the association of a policy changes with subsequent outcomes (eg, number of days until next fill following new prior authorization requirements) comparing changes in the intervention group with changes in a comparison group that was not exposed to the policy change.²⁸ The event study is a quasi-experimental design similar to a difference-in-differences study but allows for treatments that occur at various times.^29,30 A key assumption for both designs is similar trends in the period before the policy change.

To isolate the effect of prior authorization from other factors potentially influencing prescription fills and adherence, we studied beneficiaries consistently filling the drug of interest (Data Supplement, Appendix A, online only) before the policy change. Because most policy changes occurred at the start of the new plan year and because out-of-pocket costs reset on January 1, we studied the 74% of drug-policy change observations that started on January 1 in a given year (index date). We included beneficiaries with least three 30-day fills in the 120 days (September-December) preceding the new prior authorization on that drug, and we examined fills in the 120 days (January-April) following the new policy. We required that patients be alive and continuously enrolled in the same plan from 120 days before through 120 days after the policy change. The control group consisted of beneficiaries meeting these same utilization and enrollment criteria, but who were enrolled in plans that did not implement a prior authorization policy change for the drug of interest (no prior authorization or existing prior authorization before and after the index date).

We used a 20% random sample of Medicare claims data from 2010 to 2020 (Part D Prescription Drug Event, Formulary, Plan Characteristics Files, and Master Beneficiary Summary File). Medicare Part D claims included traditional Medicare and Medicare Advantage enrollees. We identified oncology drugs using the Oncology Care Model triggering drug list, and we linked the American Community Survey 5-year files for each year to obtain area-level beneficiary characteristics.^31,32 We excluded small numbers of patients for whom ZIP code data were not available or who filled drugs in greater quantities than 30 days. Our pooled sample consisted of 211,123 prescription fills for 25,136 unique beneficiaries (Data Supplement, Appendix B).

Outcome Variables

Our primary outcomes of interest were discontinued medications or delayed prescription fills. Discontinuation assessed whether beneficiaries ever refilled their prescription in the first 120 days of the new plan year. Those who switched to a different drug within the same therapeutic class were considered to have continued the drug (n = 315, 1.2% of beneficiaries).

To measure delays in the time to next fill after the index date (index fill), we considered the days' supply of medication dispensed on the last observed fill date before the new plan year and calculated the date they would require a refill—the expected fill date (Fig 1). We defined time to next fill as the difference in days between the expected fill date and the observed fill date. We required at least three observed fills in the preperiod for both treated and control groups; if we did not observe a fill in the postperiod, we censored the delay at 120 days. We additionally measured delays in fills of >30 days, 60 days, or 90 days.

FIG 1.

Study design. Event date (January 1): start of new plan year for all beneficiaries and date of prior authorization policy change for treated group. Expected fill: date beneficiary is expected to need a refill on the basis of the number of days of medication supplied at previous fill. Observed fill: date of refill (if any) observed in claims.

Covariates

We controlled for beneficiary demographics including age (younger than 65, 65-69, 70-74, 75-79, 80 or older), documented sex, race and ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, other/unknown), dual enrollment in Medicare and Medicaid, and residential ZIP code level measures of educational attainment, and the percent of residents living below the poverty level. We included indicator variables for plan type (stand-alone prescription drug plan or Medicare Advantage) and year, and we controlled for beneficiaries' average time to fill in the preperiod for each beneficiary drug fill observation.

Statistical Analysis

The unit of analysis was the beneficiary. We pooled data across all years to increase sample sizes.

Discontinuation

We estimated the unadjusted and adjusted probability of discontinuation within 120 days for both treated and control groups using a generalized linear probability model with a binary outcome indicating whether any subsequent fill was observed in the 120 days after the index date. We present adjusted odds of discontinuation associated with a prior authorization. We used the same approach to estimate the adjusted odds of prolonged delays (>30, 60, and 90 days).

Delays

We estimated time to next fill in days between expected fill date (on the basis of the number of days' supply of medication dispensed at the previous fill) and fill date observed in claims (if any) after the index date using a difference-in-differences linear regression model. The difference-in-differences approach allowed us to estimate the time to next fill in days in the preperiod and postperiod for both treated (new prior authorization policy) and control (no change in prior authorization policy) groups; the difference between them is the differential impact of a new prior authorization (Data Supplement, Appendix C). Our primary analysis grouped all drugs together, and in secondary analyses, we examined delays in time to next fill for each drug individually. The Data Supplement (Appendix D) demonstrates that time to next fill for the fills during the preperiod were nearly identical in treated and control groups, suggesting no differential pretrends.

For all analyses, we clustered standard errors at the beneficiary level. Missing data were infrequent: 1,329 (1.5%) beneficiaries across all years had missing or unmatched ZIP codes and were excluded (Data Supplement, Appendix B); 314 (1.2%) beneficiaries had unknown race and ethnicity and were grouped with other. Two-sided P values < .05 were considered statistically significant. All analyses were performed using Stata 16 (StataCorp LLC, College Station, TX).

Sensitivity Analyses

As a falsification test, we repeated our main analyses for the same 120-day periods surrounding January 1 of the year before the study year (when there were no prior authorization policy changes). We estimated the same linear probability and difference-in-differences models for beneficiaries enrolled in plans subsequently identified as treated or control plans in the analysis year. In addition, we repeated analyses after sequentially dropping each drug from the group to ensure that our findings were not driven by a single drug.

RESULTS

Study Population

Pooled over 2010-2020, the study population included 2,495 beneficiaries filling a drug with a new prior authorization policy and 22,641 beneficiaries filling the same drug in the control group with no policy change (92% of these were always prior authorization; Table 1). Of the drugs with a prior authorization change, the most common were lenalidomide (37.8%) and brand imatinib (23.6%).

TABLE 1.

Study Population Characteristics

Characteristic	New Prior Authorization (n = 2,495)	Control (n = 22,641)
Age, years, No. (%)
<65	447 (17.9)	2,515 (11.1)
65-69	499 (20.0)	5,102 (22.5)
70-74	552 (22.1)	5,437 (24.0)
75-79	434 (17.4)	4,466 (19.7)
≥80	563 (22.6)	5,121 (22.6)
Sex, No. (%)
Female	1,131 (45.3)	9,331 (41.2)
Male	1,364 (54.7)	13,310 (58.8)
Race and ethnicity, No. (%)
White, non-Hispanic	1,807 (72.4)	16,930 (74.8)
Black, non-Hispanic	423 (17.0)	3,614 (16.0)
Hispanic/Latino	96 (3.9)	672 (3.0)
Other/unknown	169 (6.8)	1,425 (6.3)
Plan type, No. (%)
Stand-alone Part D	1,857 (74.4)	13,901 (61.4)
Medicare Advantage	638 (25.6)	8,740 (38.6)
Medicaid Dual Eligibility, No. (%)	785 (31.5)	5,290 (23.4)
Proportion of residents in ZIP code tabulation area below federal poverty level, mean (SD)	12.6 (8.3)	11.8 (7.5)
Proportion of residents in ZIP code tabulation area who are college graduates, mean (SD)	28.0 (16.2)	30.0 (16.8)
Drugs studied, No. (%)
All drugs	2,495 (100)	22,641 (100)
Lenalidomide	916 (37.8)	7,720 (34.1)
Imatinib, brand	573 (23.6)	1,095 (4.8)
Abiraterone	245 (10.1)	3,122 (13.8)
Erlotinib	208 (8.6)	1,899 (8.4)
Enzalutamide	124 (5.1)	3,258 (14.4)
Pomalidomide	122 (5.0)	1,338 (5.9)
Dasatinib	116 (4.8)	688 (3.0)
Nilotinib	103 (4.2)	518 (2.3)
Thalidomide	55 (2.3)	300 (1.3)
Imatinib, generic	23 (0.9)	1,056 (4.7)
Palbociclib	10 (0.4)	1,647 (7.3)

Abbreviation: SD, standard deviation.

Demographic characteristics were similar for treated and control groups. The proportion of beneficiaries enrolled in Medicare Advantage plans was smaller in the treated group (25.6%) than the control group (38.6%). The proportion of beneficiaries dually enrolled in Medicare and Medicaid was larger in the treated group (31.5%) than the control group (23.4%). The treated group had a slightly larger proportion of beneficiaries younger than 65 years than the control group and had a slightly larger proportion of women. Race and ethnicity were similar for both groups.

Medication Discontinuation

Unadjusted, 7.6% (95% CI, 6.4 to 8.7) of the exposed group and 1.5% (95% CI, 1.3 to 1.8) of the control group discontinued their medication within 120 days after the index date. The adjusted probability of discontinuation within 120 days after the index date was 5.8% for the group exposed to a new prior authorization policy and 1.4% for the control group. The adjusted odds ratio (aOR) of discontinuation was 7.1 times greater for the exposed group relative to the control group (95% CI, 6.0 to 8.5; P < .001; Table 2).

TABLE 2.

Discontinuation or Prolonged Delays Within 120 Days After New PA

Time Measure	New PA, No.	Control, No.	Unadjusted Probability of Discontinuation, % (95% CI)		Adjusted Probability of Discontinuation, % (95% CI)a		Adjusted Odds of Discontinuation (95% CI)a
			New PA	Control	New PA	Control
Discontinuation of drug within 120 days	184	520	7.6 (6.4 to 8.7)	1.5 (1.3 to 1.8)	5.8 (4.8 to 6.7)	1.4 (1.2 to 1.5)	7.1 (6.0 to 8.5)
Time Measure	New PA, No.	Control, No.	Unadjusted Probability of Delay, % (95% CI)		Adjusted Probability of Delay, % (95% CI)a		Adjusted Odds of Delay (95% CI)a
			New PA	Control	New PA	Control
Delay >30 days	443	2,996	18.1 (16.5 to 19.8)	8.2 (7.7 to 8.5)	21.7 (19.7 to 23.6)	7.1 (6.5,7.7)	2.8 (2.5 to 3.2)
Delay >60 days	283	1,501	11.6 (10.3 to 13.0)	3.9 (3.5 to 4.2)	15.2 (13.5 to 16.9)	5.5 (5.0 to 6.0)	2.3 (1.9 to 2.6)
Delay >90 days	222	851	9.1 (7.8 to 10.3)	2.3 (2.1 to 2.6)	12.4 (10.8 to 14.1)	5.1 (4.6 to 5.7)	1.9 (1.6 to 2.2)

Abbreviation: PA, prior authorization.

^a P < .001 for all estimates.

Delayed Fills

In the fills before the index date, the average difference between expected and observed time to fill was 2.3 days in the group exposed to a new prior authorization and 2.2 days in the control group, adjusted for patient and plan characteristics (Fig 2). Thus, for a 30-day supply of medications, the expected time to next fill was 32.3 and 32.2 days, respectively. The fill after the index date was the first fill of the new plan year. For the control group—beneficiaries whose prescriptions did not have a prior authorization policy change—the adjusted time to fill was 38.7 days. For the exposed group, introduction of a new prior authorization policy was associated with an adjusted time to next fill of 48.6 days, 16.3 days later than the preperiod average. The differential change in time to fill between treated and control groups was 9.7 days (95% CI, 8.2 to 11.2; P < .001; Fig 2, Table 3). Patterns were similar for each drug subgroup (Table 3).

FIG 2.

Adjusted difference-in-differences for time to next fill in days, all drugs. *** P < .001.

TABLE 3.

DiD: Delay (in days) of Time-to-Fill After New PA Policy

Expected v Actual Fill, Delay, Days	New PA, No.	Control, No.	New PA		Control		DiD Estimate of Delay, Days	95% CI
			Delay, Preperiod, Days	Delay, After Policy, Days	Delay, Preperiod, Days	Delay, After Policy, Days
All drugs	2,495	22,641	2.3	18.6	2.2	8.7	9.7	8.2 to 11.2
Lenalidomide	916	7,720	2.9	25.2	3.3	10.6	15.0	12.0 to 17.9
Imatinib, brand	573	1,095	2.8	15.8	2.5	11.2	4.9	1.9 to 8.0
Abiraterone	245	3,122	0.5	8.8	0.5	5.5	3.4	0.1 to 6.8
Erlotinib	208	1,899	2.2	16.7	0.8	5.9	9.4	4.5 to 14.4
Enzalutamide	124	3,258	1.2	10.4	1.2	7.0	3.4	–1.9 to 8.8
Pomalidomide	122	1,338	2.6	13.0	3.1	10.7	2.8	–3.2 to 8.7
Dasatinib	116	688	3.2	20.6	2.0	9.1	10.4	3.6 to 17.1
Nilotinib	103	518	2.9	18.0	3.7	11.4	7.5	1.0 to 13.9
Thalidomide	55	300	2.5	12.9	2.1	11.8	0.7	–6.5 to 8.1
Imatinib, generic	23	1,056	0.4	23.9	1.6	8.1	16.9	–2.7 to 36.5
Palbociclib	10	1,647	1.0	10.5	1.6	6.8	4.2	–3.0 to 11.5

Abbreviations: DiD, difference-in-differences; PA, prior authorization.

The adjusted probability of prolonged delay varied by length of delay. The adjusted probability of delay >30 days after the policy change was 21.7% and 7.1% for the exposed and control groups, respectively (aOR, 2.8 [95% CI, 2.5 to 3.2]; P < .001; Table 2). The adjusted odds of a delay >60 days was 2.3 (95% CI, 1.9 to 2.6; P < .001), and the adjusted odds of a delay >90 days was 1.9 (95% CI, 1.6 to 2.2; P < .001).

Factors Associated With Treatment Delays

Several covariates had a statistically significant association with time to fill (Table 4). Beneficiaries age younger than 65 years had a 0.8-day later fill (95% CI, 0.4 to 1.1; P < .001) than beneficiaries age 65-69 years. Compared with men, women had a 0.8-day later fill (95% CI, 0.4 to 1.1; P < .001). Relative to White, non-Hispanic beneficiaries, the time to next fill was 0.7 days (95% CI, 0.5 to 1.0; P < .001) later for Black beneficiaries and 0.6 days (95% CI, 0.03 to 1.2; P = .039) later for Hispanic/Latino beneficiaries. Compared with stand-alone prescription drug plans, enrollment in Medicare Advantage was associated with a 0.2-day earlier fill (95% CI, –0.4 to –0.1; P ≤ .040). Medicaid dual eligibility status was not significantly associated with time to fill. A higher proportion of residents in the beneficiary's residential ZIP code below the federal poverty level was associated with longer delays but educational attainment was not.

TABLE 4.

Factors Associated With Treatment Delays

Characteristic	Difference, Days	95% CI	P
Age, years, No. (%)
<65	0.8	0.4 to 1.1	<.001
65-69	Ref
70-74	0.3	–0.2 to 0.3	.781
75-79	–0.1	–0.3 to 0.2	.561
≥80	0.1	–0.2 to 0.3	.453
Sex, No. (%)
Female	0.7	0.5 to 0.8	<.001
Male	Ref
Race and ethnicity, No. (%)
White, non-Hispanic	Ref
Black, non-Hispanic	0.7	0.5 to 1.0	<.001
Hispanic/Latino	0.6	0.03 to 1.2	.039
Other/unknown	–0.3	–0.7 to 0.0	.055
Plan type, No. (%)
Stand-alone Part D	Ref
Medicare advantage	–0.2	–0.4 to –0.01	.040
Medicaid dual eligibility, No. (%)	0.01	–0.2 to 0.2	.903
Per 10 percentage point increase in proportion of residents in ZIP code tabulation area with incomes below federal poverty level	2.5	1.0 to 3.9	.001
Per 10 percentage point increase in proportion of residents in ZIP code tabulation area who are college graduates	0.3	–0.26 to 0.9	.281
Year (change for each more recent year)	0.1	0.1 to 0.2	<.001

Sensitivity Analyses

We conducted falsification tests using the months surrounding the first of January 1 year before the event year, a time by when there were no prior authorization policy changes in the population. Consequently, we would not expect differences between the treated and control groups in the falsification test period. The falsification groups consisted of beneficiaries in plans that later became treated or control plans in the event year. The adjusted probability of discontinuation within 120 days after the falsification index date was 1.8 (95% CI, 0.01 to 3.5) for patients in plans that ultimately implemented prior authorization and 1.3 (95% CI, 0.1 to 1.9) in control plans. The adjusted odds of discontinuation was not statistically different (aOR, 1.8 [95% CI, 0.7 to 4.9]; P = .24; Data Supplement, Appendix E). The difference-in-differences in days delayed in the falsification year was greatly attenuated and statistically nonsignificant (1.6 days [95% CI, –3.0 to 6.2]; P = .50; Data Supplement, Appendix F). The magnitude and significance of the main result was not sensitive to exclusions of any of the 11 drug subgroups (not shown).

DISCUSSION

We found that for Medicare Part D beneficiaries consistently filling an oral anticancer drug, the introduction of a new prior authorization policy on that drug increased the odds of discontinuation within 120 days and was associated with an average delay of 10 days in refilling the first prescription after the policy change, compared with Medicare beneficiaries whose plans did not change the prior authorization policy on those drugs. Because prior authorization is so prevalent among oral anticancer drugs, more than 90% of the group without a policy change were subject to prior authorization throughout the study period. This means the results reflect the difference between new and preexisting prior authorizations rather than no prior authorization at all.

The clinical implications of discontinuations or delayed fills likely vary by drug. Gaps in adherence to oral anticancer drugs are concerning because of the risk of acquired drug resistance.²⁵ For some patients with aggressive or advanced disease, treatment gaps may precipitate disease progression.²⁴ Oral anticancer drugs have enabled some conditions to be well-controlled over a period of years, and recent guidelines outline circumstances in which treatment holidays or discontinuations may be appropriate under careful monitoring.^33-35 Nevertheless, planned treatment holidays are different than unexpected interruptions.

An important consideration when interpreting the impact of prior authorization is the efficacy of the drug—overall and for a particular patient. This study focused on established, effective treatments for which prior authorization may have limited advantages and notable disadvantages, as we observed. However, there may be some situations in which prior authorization has a role. Oncology drugs have increasingly been approved with provisional evidence of efficacy which often remain on the market after failure of confirmatory trials.^36-38 When drug approval processes do not clarify a drug's therapeutic value, this task is deferred to point-of-care utilization management processes.³⁹ The ongoing presence of unproven or disproven drugs in the market suggests that there are important and necessary applications of prior authorization to protect patients from costly, low-value care.

Prior authorization is gaining momentum as a policy issue, motivated by concerns about administrative burden and access to care.^40-42 Several states have passed legislation requiring electronic prior authorization processes, and the Centers for Medicare and Medicaid Services has issued a proposed rule that would require electronic prior authorization and other process improvements for publicly funded insurance programs including Medicare Advantage, Medicaid, and Affordable Care Act plans.⁴³ Administrative burden is at least in part a feature rather than a bug of prior authorization, but process improvements such as standardized, electronic forms are an important policy priority to alleviate the burden on a stressed health care workforce and patients. The role of prior authorization as a mechanism for determining the value of a drug is more complex. Although prior authorization may be an inefficient, frustrating approach to making decisions about resource allocation, in the current policy landscape, reducing prior authorization and reducing costs are likely to be conflicting priorities.

Our results suggest concerns about delayed and foregone care related to prior authorization are warranted. We found prior authorization primarily served to introduce delays into established drug regimens which most patients ultimately resumed. On the basis of this study context, prior authorization wasted time and undermined the policy priorities of access to care and oral anticancer drug adherence for patients who were regular users of a particular medication. However, as noted above, there are other clinical contexts where utilization review may be a safeguard against the provision of low-value care. Financial toxicity remains a serious barrier to patients' accessing oral anticancer drugs.^8,9 Thus, there may be some situations, such as decisions about initiating a new low-value drug, in which prior authorization may benefit patients. There are no easy one-size-fits-all policy solutions to prior authorization reform. Preventing new prior authorizations on established treatments, as in our study, makes sense and is a focus of several current proposals. But for less proven treatments, the impulse to remove utilization review needs to be balanced with the potential consequences for spending and quality.^20,44-46

This study had several limitations. We could not observe diagnoses or indications for prescriptions in the Medicare Part D files to ensure that the treatments were clinically indicated; however, we restricted our analysis to beneficiaries consistently filling a prescription for the given drug in the months before the policy change under the assumption that established treatment regimens suggested the drug was clinically effective and that patients could consistently access the drug. Future studies should assess the impact of such policies on clinical outcomes. Second, this study design precluded us from examining new drug starts, which may be associated with more complex prior authorization processes that may have larger effects on access. Third, this analysis examined 11 oral oncology drugs in Medicare Part D, and further study is warranted to explore the influence of prior authorization in other drugs and patient populations. Finally, most of the new-onset prior authorizations were observed in 2010-2015. The rising prevalence of prior authorization in oral oncology drugs limited our ability to study its effects—the near universality of prior authorization in recently approved oral oncology drugs left no control group to compare them.⁴⁷

In conclusion, this study found that new prior authorization policies were associated with discontinued and delayed care. Waiving prior authorization for patients already established on a drug may improve adherence.

AUTHOR CONTRIBUTIONS

Conception and design: All authors

Collection and assembly of data: All authors

Data analysis and interpretation: All authors

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Prior Authorization and Association With Delayed or Discontinued Prescription Fills

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jco/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).