Influence of policy, regulation, and insurance on clinical decision making.

VOL. 29, NO. 7 331THE AMERICAN JOURNAL OF MANAGED CARE

P rior authorization (PA) of prescription medications is used

by health insurers to manage access to costly and/or low-

value medications to ensure safe, effective, and value-based

use of medications.1 However, meta-analyses have found that PA

and other formulary restrictions can adversely influence medica-

tion adherence, clinical outcomes, and treatment satisfaction.2,3

Moreover, recent provider surveys have affirmed that medication PA

disrupts workflow and affects provider satisfaction. For instance, an

American Medical Association survey4 reported that providers spend

a mean of 14.4 hours on PAs per week, and a significant majority of

providers rated the burden of insurance company administrative

challenges as “high” or “extremely high.” Almost two-thirds stated

that they wait at least a day to hear from the insurance company

on a PA request and slightly under one-third reported that they

wait at least 3 business days. The type of medical specialty can also

significantly contribute to the PA burden, as noted in the 2019 ePA

National Adoption Scorecard published by CoverMyMeds. These

findings have been reinforced by more in-depth qualitative studies

highlighting provider burden due to extensive paperwork and

inconsistent PA requirements among health plans.5-7

There is also growing evidence that the medication PA process

can delay patient care and contribute to adverse patient outcomes.8-13

When surveyed, 24% of physicians noted that PA had led to a serious

adverse event in at least 1 of their patients, 91% reported that PA

requirements caused care delays, and 90% reported that PA delays

resulted in worse patient outcomes.4 Furthermore, PA-related cost

restrictions create delays that result in patients abandoning prescrip-

tions.14 Given these consequences and the burden of PA, providers,

pharmacists, policymakers, and other stakeholders have supported

efforts to limit, standardize, and/or streamline PA processes.15-18

Although there is a body of evidence on the benefits and unintended

consequences of PA, there is a gap in the literature on whether

medication PA alters providers’ clinical decision-making and use

of work-arounds to avoid PA requirements. We identified 1 study

in this area, which was a survey of psychiatrists.19 In that survey, a

majority of providers reported at least occasionally using tactics such

as diagnosis modification or falsification of previous medication

Influence of Prior Authorization Requirements on Provider Clinical Decision-Making Stephen G. Salzbrenner, MD; Maxwell Lydiatt, MD; Brandon Helding, PhD; Lawrence M. Scheier, PhD; Harrison Greene, MD; Patricia Wonch Hill, PhD; and Carrie McAdam-Marx, PhD

ABSTRACT

OBJECTIVES: Prior authorization (PA) aims to promote the safe and effective use of medications and to control costs. However, PA-related administrative tasks can contribute to burden on health care providers. This study examines how such tasks affect treatment decisions.

STUDY DESIGN: Cross-sectional, online survey.

METHODS: We conducted an online survey of US medical providers in 2020 based on a convenience sample of 100,000 providers. Multivariate path analysis was used to examine associations among provider practice characteristics, step therapy and other health plan requirements, perceived burdens of PA, communication issues with insurers, and prescribing behaviors (prescribing a different medication than planned, avoiding prescribing of newer medications even if evidence-based guideline recommendations are met, and modifying a diagnosis). Weighted analyses were conducted to assess nonresponse bias.

RESULTS: A total of 1173 respondents (1.2% response rate) provided 1147 usable surveys. Step therapy requirements had the largest effect on clinical decision-making. Other significant effects on clinical decision-making included perceived PA likelihood, communication issues, and health plan requirements (eg, clinical documentation). Weighted analyses showed that the study conclusions were unlikely to have been biased by nonresponse.

CONCLUSIONS: Respondents report that they may alter clinical decisions to avoid PA requirements and related burdens, even in cases in which use of the PA medication was clinically appropriate. Processes that reduce the administrative burden of PA through improved communication and transparency as well as standardized documentation may help ensure that PA more seamlessly achieves its goals of safe and effective use of medications.

Am J Manag Care. 2023;29(7):331-337. doi:10.37765/ajmc.2023.89394

CLINICAL

332 JULY 2023 www.ajmc.com

CLINICAL

trials to meet PA requirements. Further, two-thirds reported that

they occasionally refrain from prescribing a medication they would

otherwise prefer due to a known or expected PA requirement. However,

this prior study did not assess factors related to PA work-arounds

that can be attributed to the PA requirements and processes.

To address this concern, we conducted a nationwide survey of

providers to evaluate their attitudes toward medication PA and

whether PA requirements are linked to clinical decision-making in

patient care. By doing so, the current study provides insight into

how modifications to improve and streamline the PA process could

affect provider work-arounds and help ensure that the PA process

achieves the intended goal of promoting safe, high-value drug therapy.

METHODS Sampling Frame and Survey Administration

This study was part of an effort to design a digital solution for

medication PA conducted under the auspices of a National Institute

of Mental Health Small Business Technology Transfer grant. We

developed a 58-item survey, a process informed by focus groups

and stakeholder interviews. The study oversampled mental health

providers because they are the current focus of the software platform.

The remaining specialty types were selected based on provider

likelihood of encountering PA: dermatology, gastroenterology,

internal medicine, oncology, and rheumatology.20 These and other

nonpsychiatry specialties often encounter PA requirements for newly

emerged specialty medications or particularly expensive medications.

We administered the survey in October 2020 using the Qualtrics

platform. Invitation emails with a unique hyperlink were sent out

to 100,000 providers with emails drawn from a curated, nationwide

list of licensed providers. All participants provided consent prior

to taking the survey (University of Nebraska Institutional Review

Board No. 00000672).

Survey Instrument

Exogenous measures. A single item assessed challenges associated

with identifying step therapy requirements prior to prescribing (“In

general, how challenging is it to identify appropriate step therapy

requirements prior to prescribing a medication?”), with responses

on a 5-point scale ranging from “not at all chal-

lenging” to “extremely challenging.” A single

item assessed active patient load (“What is

your active patient load?”), with responses

on a 5-point scale ranging from “fewer than

25 patients” to “more than 200 patients.”

Nine items assessing interactions with

health plans were used to create 2 subscales,

one capturing health plan requirements related

to PA and a second reflecting communica-

tion issues that arise between provider and

health plans (see eAppendix [available at

ajmc.com] for additional details). The health

plan requirements subscale included 5 items, each starting with

“Regarding insurance companies, how often do you find them…”

and ending with the following: “requiring manual PA as opposed

to electronic”; “requiring PA for generic medications”; “denying PA

because of missing description of adverse effects”; and “denying

PA because of dosing issues or formulation issues” (separately)

(λ avg

= 0.64). Response options ranged from 1, “very rarely or never,”

to 5, “extremely often.”

The communication subscale included a unit-weighted risk

index for practice characteristics based on 4 items assessing

communication with health plans (see eAppendix for details).

These questions each started with “How often do you find that…”

and ended with the following: “it is necessary to send additional

clinical documentation,” “your software says a medication is denied

by insurance and you later discover it was approved,” “you are not

notified of a medication approval,” and “you are not notified of a

medication denial” (λ avg

= 0.65). The same 5-point response format

was used. These subscales were moderately correlated (r = 0.61;

P < .001), suggesting that the 2 scales share some variance, yet they

capture different elements of transacting with health plans.

We also created a unit-weighted risk index based on 4 items

assessing practice characteristics that included, for the provider

personally, the number of medication PAs that they completed in

a week (6-point response scale ranging from “none” to “more than

50”); how many hours per week they spent completing PAs (5-point

scale ranging from “less than 5 hours” to “more than 20 hours”);

length of time they waited for PA decision from health plan (7-point

scale ranging from “under 1 hour” to “more than 5 business days”);

and length of time, from start to finish, they typically needed to

complete PAs, including chart reviews and patient research (6-point

scale ranging from “0-3 hours” to “more than 2 weeks”). Higher

scores on the risk index reflect more perceived burdens of PA based

on practice experiences.

Endogenous measures (outcomes). Respondents were asked in

what percentage of cases they prescribe a different medication

than initially planned due to PA delays (5-point scale ranging from

“less than 10%” to “more than 50%”). A second item assessed how

often prescribers avoided prescribing newer medications due to

anticipated need for PA, even if the patients meet evidence-based

TAKEAWAY POINTS

We conducted a cross-sectional, online survey of licensed providers in the United States to evaluate attitudes toward medication prior authorization (PA) and to assess whether PA requirements are linked to clinical decisions in patient care.

› PA aims to promote appropriate use of medications while containing cost, but it places administrative burden on providers and payers.

› Providers report that they may prescribe different medications than originally planned, avoid prescribing newer medications, or modify a diagnosis to avoid PA burdens.

› Reducing administrative burden of PA through improved communication, standardization, and technology may help ensure that PA more seamlessly achieves its goals of safe and effective use of medications.

VOL. 29, NO. 7 333THE AMERICAN JOURNAL OF MANAGED CARE

Prior Authorization Impact on Decision-Making

guidelines, and a third item assessed how often respondents have

modified a diagnosis to obtain PA (5-point response scale for both

ranging from “very rarely” to “extremely often”).

Covariates. Additional information collected from respondents

included gender, race, practice specialty, and practitioner type.

Statistical Analysis

We examined relations among gender, race (White vs other), practice

specialty (psychiatrist vs other), and practitioner type (MD/DO vs

other) using χ2 tests of independence. We used the Student t test

to examine mean differences for continuous measures based on

gender, practice specialty, and practitioner type. A path regression

model was conducted using the Mplus statistical software program

(Muthén & Muthén) with maximum likelihood estimation to examine

associations between exogenous variables and PA-related changes

in clinical decision-making.21 Model fit was evaluated based on the

χ2 statistic (ratio of χ2:df < 5 equals good fit), comparative fit index22

and Tucker-Lewis index23,24 (numbers closer to 1.0 indicate better

fit), and standardized root mean residual and the root mean error

of approximation,25 both of which consider values less than 0.05 to

indicate good fit.26

Adjusting for Nonresponse

To account for sampling bias, we used propensity weighting strate-

gies27 to make the sample distributions better conform to the known

characteristics of total sampled population.20 These methods are

detailed in the eAppendix.

RESULTS A total of 1173 respondents (1.2% response rate) generated 1147 usable

surveys (at least 50% complete). The final sample was 49.6% female

with a mean (SD) age of 50.5 (12.9) years (Table 1). A majority were MD/

DO providers (76%; 69% MD and 7% DO), with smaller percentages of

nurse practitioners (14%) and physician assistants (10%). A majority

of the sample was White (67%), followed by Asian (18%). The largest

proportion of survey respondents were in psychiatry, per the study

design (44.9%), followed by internal medicine (18.0%), dermatology

(13.1%), gastroenterology (8.0%), neurology (5.9%), oncology (5.8%),

and rheumatology (4.4%).

Compared with respondents in other medical specialties, psychiatry

respondents were older, had fewer providers in their practice, had

smaller practice loads, encountered fewer medication PAs per

week, spent fewer hours doing PAs, had lower waiting time for PA

results, took less time to complete medication PAs, reported more

communication issues with health plans, and perceived lower

practice burden risk associated with PAs.

Regarding the 3 endogenous outcomes, 36% reported changing

medications due to delays in medication more than 30% of the

time, 26% reported modifying diagnoses to make the PA process

TABLE 1. Mean (SD) Comparisons Among Gender, Specialty Type, and Provider Typea

Measure All

Gender Specialty type Provider type

Men Women Psychiatry Other MD/DO Other

Exogenous measures

Age in years 50.52 (12.90) 52.65* (13.78) 48.65 (11.79) 52.56* (13.78) 48.83 (11.87) 50.67 (13.40) 50.05 (11.21)

Practice size 2.30 (1.52) 2.34 (1.53) 2.32 (1.52) 2.16* (1.48) 2.42 (1.53) 2.37* (1.56) 2.06 (1.36)

Patient load 3.76 (1.44) 3.87* (1.43) 3.72 (1.42) 3.61* (1.40) 3.88 (1.45) 3.78 (1.43) 3.70 (1.45)

PA burden risk indexb 1.32 (1.17) 1.34 (1.15) 1.32 (1.17) 0.99* (1.09) 1.59 (1.16) 1.33 (1.19) 1.29 (1.11)

PA weekly load 2.31 (1.11) 2.41* (1.18) 2.22 (1.05) 2.18* (0.97) 2.41 (1.20) 2.33 (1.13) 2.22 (1.04)

PA time weekly 1.23 (0.61) 1.26 (0.62) 1.22 (0.60) 1.15* (0.43) 1.30 (0.71) 1.25* (0.64) 1.17 (0.48)

Waiting time for PAs 4.71 (1.59) 4.61 (1.57) 4.77 (1.61) 4.46* (1.55) 4.92 (1.58) 4.70 (1.60) 4.73 (1.54)

Length of time to complete PAs 2.35 (1.68) 2.25 (1.63) 2.43 (1.73) 1.97* (1.56) 2.66 (1.72) 2.37 (1.70) 2.29 (1.62)

% of denied PAs that are then approved 3.13 (1.53) 3.10 (1.54) 3.13 (1.51) 3.20 (1.58) 3.08 (1.49) 3.17 (1.54) 3.01 (1.50)

Challenge identifying proper step therapy 2.93 (1.15) 2.99 (1.15) 2.86 (1.13) 2.88 (1.15) 2.96 (1.15) 2.99* (1.17) 2.72 (1.04)

Payer requirements —c (0.90) –0.04 (0.91) 0.04 (0.90) –0.02 (0.92) 0.01 (0.90) –0.01 (0.92) 0.02 (0.86)

Payer communication —c (0.92) –0.03 (0.92) 0.02 (0.92) –0.13* (0.93) 0.10 (0.91) –0.01 (0.95) 0.05 (0.84)

Endogenous/outcomes measures

% cases prescribe different medication 2.92 (1.37) 2.95 (1.38) 2.92 (1.36) 2.72* (1.37) 3.08 (1.35) 2.89 (1.38) 3.01 (1.33)

Frequency avoiding newer medication 3.19 (1.14) 3.16 (1.20) 3.20 (1.09) 3.28* (1.17) 3.11 (1.11) 3.17 (1.15) 3.25 (1.10)

Frequency modified diagnosis 1.95 (0.98) 1.97 (1.03) 1.92 (0.94) 1.89* (0.96) 2.00 (1.00) 1.95 (1.00) 1.95 (0.95)

PA, prior authorization. aMeans comparing column on the left to the column immediately to the right with an asterisk differ significantly at P < .05 (ie, men vs women, psychiatry vs other, and MD/DO vs other). bUnit-weighted risk index with 4 measures ranging from 0 to 4. cNot applicable as this was a weighted (standardized to mean = 0) factor score.

334 JULY 2023 www.ajmc.com

CLINICAL

easier, and 75% reported often avoiding new medications, even if

evidence-based, to avoid potential difficulties with PA. By specialty,

psychiatrists reported a lower percentage of cases in which they

prescribed a different medication, but a higher percentage in which

they avoided prescribing newer medications, even if evidence-based,

and reported fewer cases in which they would modify a diagnosis.

The Figure presents the results of the path analysis. Parameter

estimates and their CIs are contained in Table 2. After removing

paths between exogenous and endogenous variables that were not

significant, the 2 largest effects overall were identified with the paths

from step therapy to “avoid prescribing newer medication” (β = 0.24;

95% CI, 0.188-0.301; P < .001) and from step therapy to “prescribe

different medication” (β = 0.22; 95% CI, 0.166-0.279; P < .001). The

next largest effects were from health plan insurance requirements to

“avoid prescribing a newer medication” (β = 0.15; 95% CI, 0.074-0.221;

P < .001) and then from the practice burden risk index to “modify

a diagnosis” (β = 0.14; 95% CI, 0.086-0.204; P < .001). Overall, step

therapy requirements and health plan insurance requirements were

significantly related to all 3 endogenous outcomes. The remaining

exogenous measures each had 2 significant paths to the endogenous

outcomes. Although significant, the smallest effect overall was from

patient load to “prescribe different medication” (β = 0.06; 95% CI,

0.014-0.112; P < .05).

The curved double-headed arrows on the left side of the Figure

represent associations among the exogenous measures; those

on the right side represent correlations among the endogenous

outcome measures net of prediction. For the exogenous measures,

there was a moderate association between health plan require-

ments and communication issues (r = 0.65; P < .001). Associations

among the 3 outcome measures indicate that respondents linked

“avoid prescribing newer medications” with “prescribe different

medication” (r = 0.45; P < .001). The variance accounted for in

FIGURE. Final Path Model Predicting Provider Work-aroundsa

aBoxes represent measured (observed) variables. Straight lines with single-headed arrows are path regression coefficients (standardized). The curved double-headed arrows on the left-hand side represent associations among the exogenous measures; on the right-hand side, they represent correlations among the endogenous out- come measures net of prediction. Measures of health plan requirements and communication issues are factor composites. Model is trimmed of nonsignificant paths.

0.25

0.11 0.29

0.20

0.33

0.65

0.06

0.14

0.32

0.35

0.10

0.09

0.15

0.11

0.06

0.11

0.12

0.25

0.22

0.15

0.09

0.16

0.45

0.09 Avoid newer medication

Modify diagnosis

Different medication

Risk index

Step therapy

Patient load

Health plan requirements

Health plan communication

issues

VOL. 29, NO. 7 335THE AMERICAN JOURNAL OF MANAGED CARE

Prior Authorization Impact on Decision-Making

the endogenous outcomes was R2 = 0.161 for “prescribe different

medication,” R2 = 0.133 for “avoid newer medication,” and R2 = 0.103

for “modify diagnosis.” Additional model comparisons based on

gender and provider function (MD/DO vs all others) indicated

some small albeit significant differences in the model parameters.

However, upon closer inspection, these differences were trivial

and not clinically meaningful (results of these specific analyses

are available from first author).

The comparison of psychiatrists (n = 515) with other specialty

groups (n = 631) revealed a few minor differences in the magnitude

of path regression coefficients. However, the absolute value of the

differences in parameter estimates contrasting psychiatrists and

other specialty groups was small (0.002), indicating relatively similar

effects for the 2 groups. Further analyses suggest that the 2 groups

can be considered randomly drawn from the same population.

Detailed information on path model differences by specialty and

related statistical analyses are detailed in the eAppendix.

Adjusting for Nonresponse

The results of the weighted vs unweighted analyses for the fully

saturated model using the composite weight indicated some

trivial differences in model parameter estimates. However, the

overall mean difference (risk bias) in the regression parameters

was Δβs = 0.0005 and the same for the ΔSEs was 0.004. Also worth

noting, the mean difference in mean estimates for model variables

between unweighted and weighted data was 0.001 (SE = 0.006).

Taken together, the findings from the weighted and unweighted

analyses suggest that nonresponse bias had minimal effect on

the final results.

DISCUSSION To our knowledge, this study is the first to provide clarity on factors

that contribute to providers’ perceptions of medication PA burdens

and the effect of these burdens on their prescribing behavior. Other

nationwide surveys have reported that providers find PA burdensome,

both personally and administratively, and that PA affects clinical

decision-making. However, the current study drilled deeper into

what these burdens involve and how these burdens affect clinical

decision-making.

Overall, step therapy requirements had the largest effect on

prescribing behavior, being significantly associated with all

3 prescribing behavior outcomes and most strongly with prescribing

a different medication and avoiding prescribing a newer medication.

This finding is not unexpected given that step therapy is an ordered

protocol. In its most efficient form, information exchange for step

therapy is based on available medication dispensing data and is fully

automated. However, at times, step therapy can require providers

to complete extensive documentation of prior treatment history,

including adverse effects, intolerances, or medication failures.

Health plans vary not only in their PA policies but also in the

specific documentation that they require. This adds a layer of admin-

istrative burden to providers who must keep up with the different

requests from different health plans. Accordingly, we distinguished

step therapy from other measures because it represents a different

facet of provider-payer interaction specific to demonstrating a

patient’s outcomes with prior treatment. Indeed, previous studies’

findings have shown considerable variation with respect to how

providers view step therapy, specifically whether they feel it is a

viable method to ensure clinically appropriate use of medication

and for managing drug costs.28

Perceived practice burden and health plan communication issues

were both associated with prescribing behavior and changing a

diagnosis. Workplace perception of burdens was reflected in the PA

volume and time spent on completing PAs or waiting from a response

from health plans. The additional hours that accrue from processing

PAs most likely represent what respondents view as “burdens” that

have been noted in various professional surveys. Practices have tried

to ease the PA burden through numerous methods, including hiring

dedicated staff or outsourcing PA processes. In this study, we were

able to show that these burdens are linked with changes in prescribing

behavior. Interestingly, patient load, which reflects practice size,

TABLE 2. Associations Between Practice Characteristics and Perceived Burdens for PA and Provider Prescribing Behaviorsa

Structural path coefficient

Measure Prescribe different medication Avoid newer medication Modify diagnosis

Statistics β 95% CI β 95% CI β 95% CI

Perceived practice burden risk index 0.07* 0.011-0.129 −0.04 −0.101 to 0.019 0.13*** 0.069-0.190

Step therapy requirements 0.22*** 0.166-0.279 0.24*** 0.188-0.302 0.11*** 0.054-0.173

Patient load 0.08** 0.028-0.137 0.04 −0.013 to 0.098 0.06m −0.001 to 0.112

Health plan requirements 0.11** 0.040-0.188 0.15*** 0.079-0.228 0.08* 0.009-0.162

Health plan communication issues 0.11** 0.035-0.181 0.08* 0.002-0.149 0.10** 0.029-0.178

β, standardized coefficient; PA, prior authorization.

*P < .05; **P < .01; ***P < .001; m = marginal effect P < .06. aModel parameters are for fully saturated model not trimmed for nonsignificant effects. Negative signs indicate suppression in light of the positive zero-order associations among these measures. Outcome questions: “In what percentage of cases do you prescribe a different medication than initially planned due to PA delays?”; “How often do you avoid prescribing newer medications due to anticipated difficulties with PA, even if you feel patients meet evidence-based guidelines for their use?”; “How often have you modified a diagnosis to obtain a PA?”

336 JULY 2023 www.ajmc.com

CLINICAL

had a very small effect on prescribing a different medication and no

effect on the other 2 markers of clinical decision-making. These null

findings may indicate that respondents are not hampered by the size

of their practice or how many patients they see on a weekly basis as

much as by the insurance-specific requirements of medication PA.

Health plan requirements, such as the need to send additional

information to the health plan, and communication issues, such

as discrepancies between the PA decision communicated to the

provider and how it is adjudicated at the pharmacy, also signifi-

cantly influenced clinical decision-making. PA requirements and

communication issues reduce patient care time and therefore

lead respondents to seek “work-arounds” to avoid the PA process.

This study also identified modest differences between psychiatry

and other specialties. Psychiatry respondents were less inclined to

prescribe a different medication or avoid a newer medication in

response to patient load and health plan requirements compared

with other specialty groups. However, psychiatry respondents were

more likely than other specialties to avoid a newer medication when

faced with health plan communication issues. This may suggest

that psychiatry respondents alter their medication prescribing

behavior when they feel that communication with the health plan

is too demanding. This occurs despite their having smaller patient

loads and completing fewer medication PAs on average. Other

differences between psychiatry respondents and other specialty

groups may reflect the types of drugs being prescribed and the

diagnoses relevant to the different specialty types.

Study findings related to PA requirements and communication

issues highlight the fundamental desirability for standardization

of PA criteria across health plans and continued development

and adoption of digital solutions for PA processes. By reducing

PA burdens, these efforts may reduce the need for respondents to

seek work-arounds, which can lead to suboptimal drug treatment

and poor patient outcomes. Yet in the absence of widespread

solutions to the burden of PA, various stakeholders are advocating

for local and federal legislation that will limit the scope of PA and

step edits. In addition to numerous PA legislative initiatives by

individual states, the US House of Representatives recently passed

the Improving Seniors’ Timely Access to Care Act of 2021 (HR 3173),

which establishes several requirements and standards relating to

PA processes under Medicare Advantage plans. This trend is likely

to continue should the burden of PA on providers and patients

not be addressed through standardization, digital solutions, or

other means.29

Limitations

This study has certain limitations. Selection bias could be present

because providers who have a strong opinion or complaint about

the PA process might be more likely to respond. Cross-sectional

data can be used to identify association but not causation. Survey

response was low, and although expected given the online approach,

it contributes to the risk of nonresponse bias. Despite this low

response rate, the resulting sample demographic profile matched

well to national numbers for all physicians. Although propensity

weighting procedures were used to bring the sample data in line

with population estimates, the pool of adjustment variables was

limited. As a result, the final sample may contain moderate bias,

which may affect generalizability of the findings.

We did not sample all medical specialties but focused on providers

with high-volume medication PA, according to the CoverMyMeds

2019 ePA National Adoption Scorecard. Future studies may want

to include additional specialties with different PA volume and

experiences and widen the subject pool to include pharmacists,

administrators, and support staff involved in the PA process. Finally,

we did not qualify PA process burdens by payer type. Given Medicare

Part D requirements for expedited processing of medication PAs,

future studies may want to consider payer type and delve more

deeply into how communication with health plans can be modified

to reflect patient-centric concerns.

CONCLUSIONS In this nationwide survey, respondents reported that they may alter

clinical decisions to avoid PA requirements and related burdens,

even in cases in which use of the PA medication was clinically

appropriate. Processes that reduce the administrative burden of

PA through improved communication and transparency as well as

standardized documentation may help ensure that PA more seam-

lessly achieves its goals of safe and effective use of medications. n

Author Affiliations: Department of Psychiatry (SGS, ML, HG) and Department of Pharmacy Practice and Science (CM-M), University of Nebraska Medical Center, Omaha, NE; Methodology Evaluation Research Core, Social & Behavioral Sciences Research Consortium, University of Nebraska, Lincoln (BH, PWH), Lincoln, NE; LARS Research Institute, Inc (LMS), Sun City, AZ; Prevention Strategies (LMS), Greensboro, NC.

Source of Funding: National Institute of Mental Health (NIMH) #1R41MH124600-01.

Author Disclosures: Dr Salzbrenner is the CEO of Breezmed LLC, an electronic prior authorization company (the University of Nebraska Medical Center Board of Regents owns the Breezmed intellectual property and licenses it to Breezmed); received grants funding this research from NIMH; and reports patents received and pending for “Healthcare Provider Interface for Treatment Option and Authorization.” Dr Lydiatt was paid via NIMH grant for the present study. Dr Helding was part of the team funded by an NIMH grant that conducted the research that led to this publication. The remaining authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (SS, ML, BH, LMS, PWH, CM-M); acquisition of data (ML, BH, LMS, PWH); analysis and interpretation of data (SS, BH, LMS, PWH, CM-M); drafting of the manuscript (SS, ML, BH, LMS, HG); critical revision of the manuscript for important intellectual content (SS, ML, BH, LMS, HG, CM-M); statistical analysis (BH, LMS, PWH); provision of patients or study materials (PWH); obtaining funding (SS, PWH); administrative, technical, or logistic support (BH, HG, PWH); and supervision (SS, LMS).

Address Correspondence to: Stephen G. Salzbrenner, MD, Department of Psychiatry, University of Nebraska Medical Center, 985578 Nebraska Medical Center, Omaha, NE 68198-5578. Email: [email protected].

REFERENCES 1. 2018-2019 Academy of Managed Care Pharmacy Professional Practice Committee. Prior authorization and utilization management concepts in managed care pharmacy. J Manag Care Spec Pharm. 2019;25(6):641-644. doi:10.18553/jmcp.2019.19069 2. Happe LE, Clark D, Holliday E, Young T. A systematic literature review assessing the directional impact of managed care formulary restrictions on medication adherence, clinical outcomes, economic outcomes, and health care resource utilization. J Manag Care Spec Pharm. 2014;20(7):677-684. doi:10.18553/jmcp.2014.20.7.677

VOL. 29, NO. 7 337THE AMERICAN JOURNAL OF MANAGED CARE

Prior Authorization Impact on Decision-Making

3. Park Y, Raza S, George A, Agrawal R, Ko J. The effect of formulary restrictions on patient and payer outcomes: a systematic literature review. J Manag Care Spec Pharm. 2017;23(8):893-901. doi:10.18553/jmcp.2017.23.8.893 4. 2019 AMA prior authorization (PA) physician survey. American Medical Association. June 2020. Accessed December 20, 2020. https://web.archive.org/web/20200625150228/https://www.ama-assn.org/system/ files/2020-06/prior-authorization-survey-2019.pdf 5. Bhattacharjee S, Murcko AC, Fair MK, Warholak TL. Medication prior authorization from the providers perspective: a prospective observational study. Res Social Adm Pharm. 2019;15(9):1138-1144. doi:10.1016/j.sapharm.2018.09.019 6. Jones LK, Ladd IG, Gregor C, Evans MA, Graham J, Gionfriddo MR. Understanding the medication prior- authorization process: a case study of patients and clinical staff from a large rural integrated health delivery system. Am J Health Syst Pharm. 2019;76(7):453-459. doi:10.1093/ajhp/zxy083 7. Kane L. Medscape national physician burnout & suicide report 2020: the generational divide. Medscape. January 15, 2020. Accessed March 8, 2022. https://www.medscape.com/slideshow/2020-lifestyle- burnout-6012460#22 8. Abouzaid S, Jutkowitz E, Foley KA, Pizzi LT, Kim E, Bates J. Economic impact of prior authorization policies for atypical antipsychotics in the treatment of schizophrenia. Popul Health Manag. 2010;13(5):247-254. doi:10.1089/pop.2009.0063 9. Bianchini ML, Kenney RM, Lentz R, Zervos M, Malhotra M, Davis SL. Discharge delays and costs associated with outpatient parenteral antimicrobial therapy for high-priced antibiotics. Clin Infect Dis. 2020;71(7):e88-e93. doi:10.1093/cid/ciz1076 10. Desai S, Lo K, Nambudiri VE, Villa-Ruiz C, LaChance A, Vleugels RA. Challenges of securing insurance approval for oral tofacitinib for the treatment of alopecia areata: a multi-institution retrospective review. Arch Dermatol Res. 2022;314(5):487-489. doi:10.1007/s00403-020-02158-y 11. Dudiak GJ, Popyack J, Grimm C, Tyson S, Solic J, Ishmael FT. Prior authorization delays biologic initiation and is associated with a risk of asthma exacerbations. Allergy Asthma Proc. 2021;42(1):65-71. doi:10.2500/aap.2021.42.200101 12. Wallace ZS, Harkness T, Fu X, Stone JH, Choi HK, Walensky RP. Treatment delays associated with prior authorization for infusible medications: a cohort study. Arthritis Care Res (Hoboken). 2020;72(11):1543-1549. doi:10.1002/acr.24062 13. Yu NY, Sio TT, Mohindra P, et al. The insurance approval process for proton beam therapy must change: prior authorization is crippling access to appropriate health care. Int J Radiat Oncol Biol Phys. 2019;104(4):737-739. doi:10.1016/j.ijrobp.2019.04.007 14. Mehta D, Davis M, Epstein AJ, Lee A. Impact of formulary restrictions on antiepileptic drug dispensation outcomes. Neurol Ther. 2020;9(2):505-519. doi:10.1007/s40120-020-00195-3

15. AMCP Electronic Prior Authorization Work Group. Proceedings of the AMCP Partnership Forum: NCPDP electronic prior authorization standards—building a managed care implementation plan. J Manag Care Spec Pharm. 2015;21(7):545-550. doi:10.18553/jmcp.2015.21.7.545 16. Hoffman S. Step therapy: legal and ethical implications of a cost-cutting measure. Case Western Research University School of Law. 2018. Accessed April 26, 2021. https://scholarlycommons.law.case.edu/ faculty_publications/2009 17. Miller AP, Shor R, Waites T, Wilson BH. Prior authorization reform for better patient care. J Am Coll Cardiol. 2018;71(17):1937-1939. doi:10.1016/j.jacc.2018.03.465 18. Nayak RK, Pearson SD. The ethics of ‘fail first’: guidelines and practical scenarios for step therapy coverage policies. Health Aff (Millwood). 2014;33(10):1779-1785. doi:10.1377/hlthaff.2014.0516 19. Barnett BS, Bodkin JA. A survey of American psychiatrists concerning medication prior authorization requirements. J Nerv Ment Dis. 2020;208(7):566-573. doi:10.1097/NMD.0000000000001171 20. Physician specialty data report: number of people per active physician by specialty, 2019. Ameri- can Association of Medical Colleges. Accessed January 14, 2022. https://www.aamc.org/what-we-do/ mission-areas/health-care/workforce-studies/interactive-data/number-people-active-physician-specialty-2019 21. Muthén LK, Muthén BO. How to use a Monte Carlo study to decide on sample size and determine power. Struct Equ Modeling. 2002;9(4):599-620. doi:10.1207/S15328007SEM0904_8 22. Bentler PM. Comparative fit indexes in structural models. Psychol Bull. 1990;107(2):238-246. doi:10.1037/0033-2909.107.2.238 23. Kaplan D. Evaluating and modifying covariance structure models: a review and recommendation. Multivariate Behav Res. 1990;25(2):137-155. doi:10.1207/s15327906mbr2502_1 24. Tucker LR, Lewis C. A reliability coefficient for maximum likelihood factor analysis. Psychometrika. 1973;38(1):1-10. doi:10.1007/BF02291170 25. Steiger JH, Lind JC. Statistically based tests for the number of common factors. Paper presented at: Psychometric Society Annual Meeting; May 28, 1980; Iowa City, IA. 26. Hu Lt, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct Equ Modeling. 1999;6(1):1-55. doi:10.1080/10705519909540118 27. Kalton G, Flores-Cervantes I. Weighting methods. J Off Stat. 2003;19(2):81-97. 28. Fischer MA, Kesselheim AS, Lu Z, Ross KM, Tessema FA, Avorn J. Physician perceptions of step therapy pre- scribing requirements. J Manag Care Spec Pharm. 2019;25(11):1210-1224. doi:10.18553/jmcp.2019.25.11.1210 29. Yang E, Yang S. Prior authorization: overwhelming burden and critical need for reform. JACC Case Rep. 2020;2(10):1466-1469. doi:10.1016/j.jaccas.2020.05.095

Visit ajmc.com/link/89394 to download PDF and eAppendix

eAppendix Table. Percentage Endorsing Each Response Level for Measures in Path Model

All Male Female Psychiatrist Other MD/DO Other

Q2 How many Advanced Practice Providers are in

your immediate practice setting?

92.40% 46.89% 52.83% 44.95% 55.05% 75.98% 24.02%

Less than 5 68.00% 32.08% 35.57% 32.62% 35.82% 52.57% 15.87%

Between 5 and 10 18.10% 8.77% 9.53% 7.00% 11.08% 12.77% 5.32%

11-15 4.70% 2.55% 2.36% 1.86% 2.84% 3.46% 1.24%

16-20 2.10% 1.04% 1.04% 0.80% 1.33% 1.86% 0.27%

20+ 6.60% 2.45% 4.34% 2.66% 3.99% 5.32% 1.33%

Q3 What is your active patient load? 99.60% 46.84% 52.88% 5.08% 5.69% 8.14% 2.63%

Under 25 10.70% 4.93% 5.30% 6.04% 7.18% 9.72% 3.50%

25-50 patients 13.20% 5.58% 7.43% 7.62% 5.87% 9.89% 3.59%

51-100 patients 13.40% 5.20% 8.09% 8.67% 5.52% 11.21% 2.98%

101-200 patients 14.10% 6.23% 7.81% 17.51% 30.82% 37.13% 11.21%

Over 200 patients 48.10% 24.91% 24.26% 44.92% 55.08% 76.09% 23.91%

Q8 How many medication prior authorizations do

you personally complete in an average week?

99.70% 47.03% 52.69% 44.79% 55.21% 76.03% 23.97%

None 21.90% 9.55% 12.34% 8.92% 13.04% 16.01% 5.95%

1-5 45.70% 20.41% 25.05% 25.20% 20.65% 35.43% 10.41%

6-10 18.60% 9.28% 9.46% 6.39% 12.25% 13.65% 4.99%

11-20 8.20% 4.73% 3.43% 2.89% 5.34% 6.47% 1.75%

21-50 3.70% 1.95% 1.95% 0.79% 2.97% 3.06% 0.70%

50+ 1.60% 1.11% 0.46% 0.61% 0.96% 1.40% 0.17%

Q9 About how many hours a week do you

personally spend completing medication prior

authorizations?

99.10% 47.11% 52.61% 44.94% 55.06% 76.17% 23.83%

Less than 5 hours 81.70% 37.69% 44.31% 39.14% 43.27% 62.01% 20.40%

5-10 hours 14.20% 7.74% 6.72% 5.19% 9.15% 11.17% 3.17%

11-15 hours 1.70% 0.93% 0.65% 0.44% 1.23% 1.67%

16-20 hours 0.60% 0.28% 0.37% 0.09% 0.53% 0.44% 0.18%

More than 20 hours 1.00% 0.47% 0.56% 0.09% 0.88% 0.88% 0.09%

Q10 On average, how long do you wait for a prior

authorization decision from health plans?

87.60% 45.73% 54.06% 46.37% 53.63% 76.22% 23.78%

Under 1 hour 5.00% 2.74% 3.16% 3.28% 2.39% 4.28% 1.39%

A few hours 6.20% 3.16% 4.00% 3.48% 3.58% 5.67% 1.39%

More than a few hours but less than 1 business day 6.00% 3.69% 2.85% 3.28% 3.58% 5.07% 1.79%

1 business day 14.60% 8.43% 8.32% 9.75% 6.97% 13.53% 3.18%

2 business days 23.50% 12.33% 14.65% 13.53% 13.33% 19.70% 7.16%

3-5 business days 25.20% 12.43% 15.91% 11.24% 17.51% 21.19% 7.56%

More than 5 business days 7.10% 2.95% 5.16% 1.79% 6.27% 6.77% 1.29%

Q11 In your experience, how long from start to

finish do prior authorizations take to complete,

including all chart reviews and patient research?

99.40% 47.08% 52.65% 44.82% 55.18% 76.14% 23.86%

0 to 3 hours 49.00% 23.96% 25.26% 28.07% 21.23% 37.81% 11.49%

4 to 8 hours 11.00% 4.83% 6.41% 4.04% 7.02% 8.68% 2.37%

1-2 days 11.90% 6.50% 5.57% 4.65% 7.28% 8.60% 3.33%

3-7 days 17.30% 7.61% 9.56% 5.88% 11.49% 12.63% 4.74%

1-2 weeks 8.50% 3.90% 4.64% 2.02% 6.49% 7.02% 1.49%

Greater than two weeks 1.80% 0.28% 1.21% 0.18% 1.67% 1.40% 0.44%

Q12 What percentage of your denied medication

requests are generally approved upon appeal?

99.70% 46.99% 52.73% 44.88% 55.12% 76.03% 23.97%

< 10% 21.40% 10.19% 11.31% 9.97% 11.55% 15.92% 5.60%

10-20% 17.60% 8.99% 8.80% 7.44% 10.24% 13.56% 4.11%

21-30% 15.50% 7.14% 8.71% 5.86% 9.71% 11.20% 4.37%

31-50% 16.50% 7.04% 9.45% 6.91% 9.62% 12.34% 4.20%

>50% 28.60% 13.62% 14.46% 14.70% 14.00% 23.01% 5.69%

Q13 In general, how challenging is it to identify

appropriate step therapy requirements prior to

prescribing a medication?

99.20% 47.02% 52.70% 45.08% 54.92% 76.19% 23.81%

Not at all challenging 9.10% 4.28% 4.84% 4.48% 4.66% 6.59% 2.55%

Somewhat challenging 31.00% 13.04% 17.69% 14.50% 16.70% 23.20% 8.00%

Moderately challenging 28.90% 14.06% 15.64% 12.92% 16.17% 20.56% 8.52%

Highly challenging 19.00% 10.15% 8.85% 8.44% 10.72% 15.99% 3.16%

Extremely challenging 11.30% 5.49% 5.68% 4.75% 6.68% 9.84% 1.58%

Q14 In what percentage of cases do you prescribe

a different medication than initially planned, due to

prior authorization delays?

99.10% 47.20% 52.52% 44.90% 55.10% 76.08% 23.92%

<10% 19.40% 9.33% 9.70% 10.99% 8.53% 15.83% 3.69%

10-20% 22.00% 9.70% 12.50% 10.73% 11.43% 16.80% 5.36%

21-30% 22.10% 10.45% 11.85% 8.88% 13.37% 15.92% 6.33%

31-50% 18.70% 9.51% 9.33% 8.27% 10.64% 14.95% 3.96%

>50% 17.00% 8.21% 9.14% 5.98% 11.17% 12.58% 4.57%

Q15 How often do you avoid prescribing newer

medications due to anticipated difficulties with

prior authorization, even if you feel patients meet

evidence-based guidelines for their use?

99.60% 46.94% 52.78% 44.83% 55.17% 76.01% 23.99%

Very rarely 8.00% 4.45% 3.80% 3.42% 4.64% 6.13% 1.93%

Rarely 18.00% 9.18% 8.44% 7.18% 10.86% 14.62% 3.42%

Often 36.10% 15.31% 21.06% 16.29% 19.96% 27.85% 8.41%

Very often 22.50% 10.20% 12.15% 9.46% 13.13% 15.32% 7.27%

Extremely often 15.00% 7.79% 7.33% 8.49% 6.57% 12.08% 2.98%

Q16 How often have you modified a diagnosis to

obtain a prior authorization?

99.30% 46.79% 52.93% 44.60% 55.40% 75.90% 24.10%

Very rarely 39.80% 18.98% 20.93% 19.05% 20.98% 30.82% 9.22%

Rarely 33.80% 15.26% 19.35% 15.36% 18.70% 25.46% 8.60%

Often 18.50% 8.56% 9.40% 7.11% 11.50% 13.87% 4.74%

Very often 5.40% 2.88% 2.51% 2.37% 3.07% 4.30% 1.14%

Extremely often 1.80% 1.12% 0.74% 0.70% 1.14% 1.49% 0.35%

Note: Numbers in columns represent percentages endorsing a particular response level. Totals may not represent all survey

respondents given missing data. Statistical comparisons based on mean levels for ordinal-categorical measures can be found in

Table 1 of the article.

Copyright of American Journal of Managed Care is the property of MJH Life Sciences and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.