# Prior Authorization of Medication and Its Influence on Provider Behavior: Latent Class Analysis

## TL;DR

This study explores how prior authorization requirements for medications affect doctors' prescribing behaviors and identifies distinct groups of providers with different experiences.

## Contribution

The study introduces a novel approach using latent class analysis to uncover unobserved subgroups of prescribers with similar PA-related behaviors.

## Key findings

- Four distinct provider subgroups were identified based on their PA-related behaviors and experiences.
- The largest group reported problematic PA experiences and was more likely to alter prescribing and diagnostic behaviors.
- Only age, specialty type, and patient load were significant in characterizing subgroup membership.

## Abstract

Insurance companies frequently require prior authorization (PA) for medication prescriptions to ensure quality control and safety. The added layer of scrutiny can contribute to provider dissatisfaction and has been associated with adverse patient outcomes. Health care providers have changed prescribing behaviors to avoid PA. Understanding factors contributing to this phenomenon can facilitate systemic change and better patient care.

The objectives of this study are to identify unique unobserved subgroups of prescribers with similar PA-related behaviors using a finite mixture modeling approach; characterize subgroup membership by important covariates; and examine the influence of subgroup membership on 3 relevant prescribing outcomes.

A cross-sectional, web-based, nationwide survey of 1173 prescribers was oversampled for psychiatry in support of developing a software-as-a-solution to facilitate PA. Latent class analysis included 12 indicators assessing the degree of PA involvement, provider-insurance communication, and the methods of obtaining or avoiding PA. Covariates included age, gender, race, provider role, specialty, number of prescribers, and patient load. Three clinical decision outcomes included prescribing medication other than initially preferred due to PA delays, avoiding newer medications due to anticipated need for PA, and modifying a diagnosis to obtain PA.

In total, 1147 prescribers responded with 1144 usable surveys (age, median 50.003 [range 25.00, 72.00] years; 569 (49.74%) females; 67.13% White; 44.84% psychiatrists). In total, 4 unique classes were obtained based on 12 indicators assessing PA-related activities. Classes included a high PA denial class (291 [25.15%]), a Low Volume PA (178 [15.93%]), a class denoted by Problematic Communication Issues with insurers (227 [19.96%]), and a Low Volume PA Class with Problematic Experiences (446 [38.97%]). Only 3 of the 7 covariates (age, specialty type, and patient load) provided additional means to characterize class membership. The observation that certain demographics (race and gender) and provider characteristics (specialty) may not be informative has policy implications and can inform means to improve provider-insurer communication. The largest class reporting problematic PA experiences had significantly higher mean levels for changing their prescribing and diagnostic behaviors than the remaining classes.

Providers are not homogeneous regarding their experience with PA and insurance companies. It is, therefore, important to recognize subtle behavioral differences and find ways to accommodate the PA process to their unique needs. This will facilitate the appropriate implementation of PA by insurance companies. Providers can then avoid the need to alter medications, change diagnoses, or resist prescribing newer, effective medications that may require lengthy clinical documentation.