Impact of a National Private Health Insurer’s Prior Authorization Policy on Utilization of Vaginal Hysterectomy

Abstract

Objectives:

On April 6, 2015, the largest private health insurer in the United States implemented a policy requiring prior authorization for all hysterectomies except those done as outpatient vaginal. The purpose of this policy was to increase utilization of vaginal hysterectomy; however, it is unknown whether this policy had its intended effect. We sought to analyze trends in hysterectomy routes before and after implementation of the prior authorization policy to see if utilization of vaginal hysterectomy increased.

Methods:

This was a retrospective study using the Optum® Clinformatics Data Mart national claims database of women enrolled in a single national private health insurer who underwent hysterectomy for any indication between January 1, 2010 and June 30, 2016. Per-quarter utilization of hysterectomy routes (abdominal, laparoscopic, vaginal, and laparoscopic-assisted vaginal) were compared between the pre- and post-policy time periods using interrupted time series analyses.

Results:

Data for 305,139 hysterectomies were available—248,821 in the pre-policy period and 56,318 in the post-period. Outpatient vaginal hysterectomy had the greatest increase in utilization of all routes and types; the average utilization per quarter in the pre-policy period was −0.61% and this increased to 0.21% in the post-policy period (p<.0001). Outpatient laparoscopic hysterectomy had the greatest decrease in utilization, with an average decrease of −1.50% per quarter.

Conclusions:

The prior authorization policy was associated with a short-term increase in utilization of vaginal hysterectomy.

Introduction

Hysterectomy is the most common major gynecologic procedure in the United States, with over 400,000 women per year undergoing this procedure.¹ The American College of Obstetricians and Gynecologists recommends the vaginal route of hysterectomy whenever feasible, as this route is associated with better outcomes, lower complication rates, and lower costs compared to other routes.² Despite evidence and recommendations supporting the use of vaginal hysterectomy, utilization of this route has been declining for the past several decades.^1,3

Private health insurers benefit financially by encouraging the use of high-value interventions. Prior studies on the effectiveness of insurance policies implemented to promote or decrease certain physician practices show mixed results.^4–6 On April 6, 2015, the largest private health insurer in the United States implemented a policy requiring prior authorization for all hysterectomies except those done as outpatient vaginal,⁷ regardless of indication. The purpose of this policy was to increase utilization of vaginal hysterectomy; however, it is unknown whether this policy had its intended effect.

Optum® Clinformatics Data Mart is a national database that contains administrative claims data for the large private insurer that implemented this vaginal hysterectomy policy. Therefore, the purpose of this study was to analyze trends in hysterectomy routes before and after implementation of the prior authorization policy to see if utilization of vaginal hysterectomy increased.

Materials and Methods

We performed a retrospective study using data from Optum® Clinformatics Data Mart, a national database of private payer claims that includes over 77 million individuals enrolled in both medical and pharmacy coverage from a national health insurer. The database also includes patients with supplemental insurance types; therefore, women with Medicare Advantage Plans may have also been included in the study cohort. The database captures all outpatient and inpatient services utilized by members throughout their enrollment, as well as outpatient pharmacy claims; complete patient payment in the form of copayments, deductibles, and coinsurance; and standardized costs (reimbursements). The study received exempt status from the University of Michigan IRB (HUM00132182). Hysterectomies performed from January 1, 2010 to June 30, 2016 were identified using Current Procedural Terminology (CPT) and International Classification of Diseases, Ninth Revision and Tenth Revision (ICD-9-CM and ICD-10-CM) codes. Benign, pre-malignant, and malignant indications were identified using ICD-9-CM and ICD-10-CM diagnosis codes at the time of hysterectomy. Hysterectomies performed for obstetrical indications and for women aged <18 years were excluded.

We identified all hysterectomies performed by quarter in both the inpatient and outpatient settings based on administrative claims. Women with outpatient hysterectomy could have spent one night in the hospital as long as the total stay was less than 24 hours. Surgical approach for hysterectomy was determined using either CPT or ICD-9-CM or ICD-10-CM procedure codes. Hysterectomies with more than one surgical approach were assigned the more invasive approach. Quarterly proportions of cases done via the abdominal, laparoscopic, laparoscopic-assisted vaginal (LAVH), and vaginal (TVH) routes were calculated in the inpatient and outpatient settings, respectively. Data regarding the robotic-assisted laparoscopic approach were not available due to lack of coding; therefore, these cases were included in the laparoscopic group.

Differences in demographic characteristics and indications for surgery for women undergoing hysterectomy in the pre- and post-policy periods were measured using effect size.⁸ Effect size is advantageous when comparisons between large groups may result in significant p-values despite little meaningful differences. A commonly used interpretation to classify the extent of the meaningful differences in effect sizes can be divided into four categories: little difference (d < 0.2), small (0.2 ≤ d < 0.5), medium (0.5 ≤ d < 0.8), and large (d ≥ 0.8).⁹ For continuous variables that were deemed normally distributed after examining skew, histograms, and Q-Q plots, we determined effect size using Cohen’s d. For categorical variables, Cohen’s h was used.

Since the vaginal hysterectomy policy was implemented on April 6, 2015, we assumed the second quarter of 2015 to represent the start of the intervention period. Therefore, the pre-policy period spanned 21 quarters, from January 2010 – March 2015, and the post-policy period was five quarters, from April 2015 – June 2016. To estimate the change in utilization of vaginal hysterectomy after policy implementation, a single-intervention multiple-group interrupted time series analysis (ITSA) was performed using Ordinary Least Squares (OLS) estimation. ITSA is a quasi-experimental design and a useful analytic approach that uses linear regression to examine the efficacy and longitudinal effects of an intervention.¹⁰ One strength of this approach is that ITSA is generally unaffected by confounding variables. By using this approach, we were able to estimate the change in utilization during the pre- and post-policy periods in the inpatient and outpatient settings, respectively, for each surgical approach.

All data extraction, cleaning, and descriptive analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA) and ITSA was performed using STATA (Stata 13/SE, College Station, TX).

Results

Data for 305,139 hysterectomies were available—248,821 in the pre-policy period and 56,318 in the post-period. Table 1 shows demographics and indications for hysterectomies performed during these two periods. The proportion of inpatient hysterectomies decreased from almost 55% in the pre-policy period to 38% in the post-policy period (small standardized difference of 0.34). All other variables had effect sizes <0.2, indicating little differences between groups.

Table 1.

Demographics comparing privately insured women who underwent hysterectomy prior to and following the insurance policy implemented April 6, 2015 that requires prior authorization for all hysterectomy routes other than outpatient vaginal

Demographics	Pre-Policy January 2010-March 2015 N=248,821	Post-Policy April 2015-June 2016 N=56,318	Effect Size
Age, years	48.35 ± 11.32	49.16 ± 11.72	0.071a
Race/Ethnicityb
Caucasian	173,672 (70.6)	35,103 (68.49)	0.046
Black/African-American	34,704 (14.11)	7,008 (13.67)	0.013
Asian	5,191 (2.11)	1,367 (2.67)	0.036
Hispanic	25,021 (10.17)	6,151 (12.0)	0.058
Unknown	7,407 (3.01)	1,625 (317)	0.009
Education Levela
Less than 12th Grade	1,053 (0.43)	306 (0.6)	0.024
High School Diploma/GED	78,934 (32.09)	16, 353 (31.91)	0.004
Some College	133,372 (54.22)	27,774 (54.19)	0.001
Bachelor Degree/Post-Graduate Degree	31,875 (12.96)	6,632 (12.94)	0.001
Unknown	761 (0.31)	189 (0.37)	0.010
U.S. Census Division
East North Central	37,681 (15.14)	8,625 (15.31)	0.005
East South Central	12,181 (4.9)	3,455 (6.13)	0.054
Middle Atlantic	10,938 (4.4)	3,063 (5.44)	0.048
Mountain	25,297 (10.17)	5,592 (9.93)	0.008
New England	6,107 (2.45)	1,039 (1.84)	0.042
Pacific	13,123 (5.27)	3,150 (5.59)	0.014
South Atlantic	70,160 (28.2)	13,468 (23.91)	0.098
Unknown	321 (0.13)	110 (0.2)	0.016
West North Central	26,785 (10.76)	6,238 (11.08)	0.010
West South Central	46,228 (18.58)	11,578 (20.56)	0.050
Household Incomea
<$40,000	43,738 (17.78)	9,490 (18.52)	0.049
$40,000 – $59,000	36,177 (14.71)	7, 581 (14.79)	0.032
$60,000 – $99,000	61,693 (25.08)	12,199 (23.8)	0.018
≥$100,000	74,477 (30.28)	14,610 (28.51)	0.016
Unknown	29,910 (12.16)	7,374 (14.39)	0.081
Inpatient	136,359 (54.8)	21,407 (38.01)	0.342
Indication for Hysterectomy
Benign	212,559 (85.43)	46,994 (83.44)	0.033
Abnormal Uterine Bleeding	115,201 (46.3)	26,145 (46.42)	0.011
Fibroids	109,641 (44.06)	26,011 (46.19)	0.010
Pelvic Organ Prolapse	38,041 (15.29)	7,623 (13.54)	0.035
Infectious Disease of Uterus/Cervix	26,392 (10.61)	5,878 (10.44)	0.008
Noninfectious Disease of Ovaries/Tubes	83,837 (33.69)	16,451 (29.21)	0.061
Pre-malignant	30,302 (12.18)	8,653 (15.36)	0.048
Cervical Dysplasia	5,895 (2.37)	1,820 (3.23)	0.029
Endometrial Hyperplasia	6,244 (2.51)	1,939 (3.44)	0.030
Other Premalignant Lesions	19,257 (7.74)	5,347 (9.49)	0.032
History of Cancer (personal or family)	9,229 (3.71)	2,454 (4.36)	0.017
Gynecologic Cancer	4,860 (1.95)	4,306 (7.65)	0.157

Data presented as mean ± SD or n/N (%).

^aEffect size determined using Cohen’s d statistic; all other effect sizes determined using Cohen’s h.

Effect size index <0.2 represents little standardized differences and ≥0.8 represents large standardized differences.

^bData unavailable for 7,890 cases for Race/Ethnicity, Education Level, and Household Income for the same surgical cases.

Figure 1 shows the interrupted time series analysis for change in utilization of inpatient and outpatient hysterectomies by surgical route. At the beginning of the study period, the most common outpatient hysterectomy route was vaginal; however, by the end of the study period, nearly 60% were performed laparoscopically. For inpatient hysterectomy, the abdominal route remained the most common route throughout the study period and the prevalence increased over time. The slope of each trend line before and after policy implementation indicates if there is decreasing or increasing utilization (Table 2). For outpatient vaginal hysterectomy, there was an average quarterly decrease of −0.61% in utilization in the pre-policy period; however, in the post-policy period, utilization increased by 0.21% per quarter and this difference was significant (p<.0001). A similar trend was seen with inpatient vaginal hysterectomy, although the overall proportion of hysterectomies performed via this route was lower compared to outpatient.

Figure 1.

Change in utilization of inpatient and outpatient hysterectomies by route before and after policy implementation

Table 2.

Average per-quarter changes in utilization of hysterectomy routes prior to and following implementation of the prior authorization policy

Route	Pre-Policy Slope % (95% CI)	Post-Policy Slope % (95% CI)	Post-Pre Difference % (95% CI)	P-value
Inpatient
Vaginal	−0.44 (−0.50, −0.37)	0.13 (−0.40, 0.56)	0.57 (0.40, 1.00)	0.0130
Laparoscopic	0.13 (−0.15, 0.41)	−0.32 (−0.68, 0.04)	−0.45 (−0.91, 0.01)	0.0570
LAVH	−0.30 (−0.35, 0.25)	−1.08 (−1.46, −0.71)	−0.78 (−1.16, −0.41)	<0.0001
Abdominal	0.61 (0.5, 0.97)	0.23 (0.02, 0.43)	−0.38 (−0.82, 0.06)	0.0860
Outpatient
Vaginal	−0.61 (−0.71, −0.50)	0.21 (0.10, 0.31)	0.82 (0.66, 0.97)	<0.0001
Laparoscopic	1.51 (1.27, 1.76)	0.01 (−0.18, 0.20)	−1.50 (−1.81, −1.19)	<0.0001
LAVH	−0.79 (−0.90, −0.67)	−0.26 (−0.40, −0.13)	0.53 (0.36, 0.01)	<0.0001
Abdominal	−0.12 (−0.19, −0.05)	0.04 (−0.01, 0.09)	0.16 (0.07, 0.26)	0.0020

LAVH: laparoscopically assisted vaginal hysterectomy

Outpatient laparoscopic hysterectomy had the largest change in utilization after policy implementation, with an average increase in utilization of 1.51% per quarter during the pre-policy period and only a 0.01% average quarterly increase during the post-policy period. Change in utilization of inpatient laparoscopic hysterectomies also shifted from an increase of 0.13% per quarter prior to the policy to a decline in utilization of −0.32% per quarter after the policy. Utilization of LAVH decreased during both time periods, with inpatient LAVH having a 3-fold decrease in rate of decline compared to the pre-policy period (pre-policy slope = −0.30 (−0.35, 0.25), post-policy slope = −1.08 (−1.46, −0.71), p-value <0.0001). Finally, utilization of inpatient abdominal hysterectomy increased by 0.61% per quarter during the pre-policy period, and while there continued to be a positive slope in the post-policy period, the rate of increase slowed to only 0.23%—a difference that was not statistically significant (p=0.086). As expected, outpatient abdominal hysterectomy was the least common hysterectomy type; however, utilization did change significantly, from a decline in the pre-policy period to an increase of 0.04% per quarter in the post-policy period.

Discussion

Implementation of a prior authorization policy for all hysterectomy routes other than outpatient vaginal resulted in a short-term increase in utilization of both inpatient and outpatient vaginal hysterectomy. The rapid uptake of laparoscopic hysterectomy also slowed in the post-policy period. This practice change took place during a time when outpatient hysterectomy became more common than inpatient hysterectomy.

This study adds to the limited body of literature addressing the impact that insurers’ policies have on medical practice patterns. The few studies that do exist on this topic show mixed results regarding the efficacy and long-term value of insurance-driven medical care. A systematic review from Dartmouth found that insurer restrictions aimed at reducing low-value care (e.g., prior authorization) can be effective at changing physician practice but may result in unintended consequences.⁵ For example, a Canadian study analyzing a policy aimed at decreasing the use of fluoroquinolones found that while the policy was successful in reducing fluoroquinolone prescriptions by over 80%, prescription rates of other antibiotics increased.¹¹ Another study by Goodman et al. analyzed the effect of a prior authorization policy aimed at reducing surgical intervention for low back pain. The policy, which mandated a referral to a physiatrist prior to referral to a spine surgeon, resulted in only a transient decrease in lumbar fusions and had the unintended consequence of increasing costs for non-operative care.¹²

Similarly, while our study did show a short-term increase in utilization of vaginal hysterectomy, it is unknown whether this will be sustained long term or whether there will be any unintended consequences. We were unable to assess intra- or post-operative complications, actual hospital costs, or other healthcare utilization measures such as length of stay, reoperations, and readmissions. However, a robust analysis looking at these factors is needed before drawing conclusions regarding the value of a policy such as this one.

Several factors concurrent with our study period may have influenced some of our results. First, there was an overall shift from inpatient to outpatient hysterectomies during the study period that pre-dated the policy. A recent analysis on hysterectomy rates in the United States found that from 2010 to 2013, outpatient hysterectomy rates increased from 13.3 per 10,000 women to 19.6 per 10,000 women.¹³ As a result, cases selected to be performed as inpatient may be more complex and associated with more healthcare utilization and greater cost, which in the current study could help explain the increased total payments for inpatient cases in the post-policy period. Secondly, the FDA warning regarding power morcellation was released on April 17, 2014—almost exactly one year prior to the hysterectomy policy.¹⁴ Several studies have demonstrated that after this warning was released, practice patterns changed, with an increase in abdominal hysterectomy and a decrease in minimally-invasive routes.¹⁵ A study by Harris et al. reported a significant increase in utilization of vaginal hysterectomy in the eight months following the FDA safety communication as compared to the prior 15 months.¹⁶ Since our study also spans this same time period, it is possible that our results may have been impacted by practice pattern changes related to morcellation concerns.

Limitations to our study include the relatively short post-policy time period and lack of data regarding complications and other relevant measures of healthcare utilization. While we did not find any large differences in our cohorts for the variables we analyzed, there may have been meaningful differences that we were unable to assess. Our study group was also limited to a single private insurer in the United States, which may not be generalizable to other groups. Strengths of our study include the large cohort and robust analytical approach using interrupted time series analysis.

In conclusion, a prior authorization policy implemented by a single national private insurer and intended to increase outpatient vaginal hysterectomy was associated with a short-term increase in utilization of outpatient and inpatient vaginal hysterectomy. Increasing utilization of high-value services should be a priority for insurers, physicians, hospitals, and patients; however, it is important to understand the impact of insurer policies meant to change physician practices. Long-term follow up is needed to determine whether the short-term increase in utilization of vaginal hysterectomy is sustained.