11606_2017_4003_MOESM1_ESM

Appendix
Price Transparency in Primary Care: Can Patients Learn About Costs When Scheduling
an Appointment?
Journal of General Internal Medicine, Saloner et al.
Appendix Figure 1. Flow Diagram
Full sample
N=7,865
Offered a price
N= 4,469
Free
N=512
Not free
N=3,957
Not offered a price
N=3,396
Referred to other
sources
N=1,885
Not referred to
other sources
N=1,511
Note: Sample size in each box indicates the number of individuals with the outcome.
1
Appendix Figure 2. Comparing Mean Prices for Physician Consultations Using Audit Data
and Insurance Claims at the Level of 3-digit ZIP Codes
We examine the extent to which the prices used in this study appear to be valid in the sense
that they measure what they were intended to, and the extent to which audit price measurement
appears to be is consistent across geographies. To do this, we examine how the prices quoted
to patients in the audit study compare to average actual payments to physicians (i.e., “allowed
amounts” or negotiated rates) in each geographic area. We obtained a comparison sample of
median prices for new patient evaluation and management physician visits for a patient with
moderate complexity (CPT 99203) at the level of 3 digit zip codes. These were provided for the
75 most common zip codes in our data (accounting for more than half of the sample data points)
under a licensing agreement from FairHealth, a national database of health claims from a broad
set of private insurance companies. We examined the relationship between average audit prices
and average median prices at the level of 3 digit zip codes and calculated measures of
association including means, regression coefficients and R-squared. The analysis is limited to
prices at 3 digit zip code level for the geographic areas represented in both datasets. In the
audit data, we limited the analysis to commercial prices (employer-sponsored private insurance
and marketplace coverage) to be most comparable to the FairHealth data.
Overall, this analysis confirms that the commercial audit prices approximately track regional
norms. The average commercial audit price at the 3 digit zip code level was similar to, but
somewhat lower than, that in the FairHealth data, $153 compared to $180. The regression
coefficient for the average actual payment amount price is 0.660 (standard error 0.123) with a yintercept of 34.800 (standard error 22.421). A moderate correlation of R=.30 between average
commercial audit prices and paid amounts within geographic areas demonstrates that
differences in prices across areas are reflected to modest degree.
There are several reasons that we would expect a difference between the average audit price
and the average actual prices paid to physicians by insurers. Offices may be quoting average
prices for a service that is different than CPT 99203 or for a commercial insurer that does not
have the average contracted rate. On average, offices quoted a price that was less than the
FairHealth price, but in many cases offices quoted a higher price than FairHealth, suggesting
some degree of inconsistency in interpretation or answering of the question. In addition, some 3
digit zip codes have a relatively large difference between the prices callers are told and the
actual prices paid, which raises a potential concern that the accuracy of price information may
differ depending on location. We also note that the differences between audit prices and actual
prices may come from differences in the provider samples.
2
Note: Each dot represents data from a single 3-digit ZIP code measured in the audit study and
in comparison data from FairHealth. Audit data reflect mean prices quoted to callers across
clinical scenarios, FairHealth data represent the median price in the area for new patient
evaluation and management physician visits for a patient with moderate complexity (CPT
99203). Regression line and 95% confidence interval is overlaid on the plot.
3
Appendix Table 1. Unconditional Ability to Obtain a Price (Treating Calls where no visit
was scheduled as zero instead of missing)
Unconditional ability to obtain
a price (N=10,065)
Estimate
95% CI
Caller characteristics
Uninsured (ref=employer)
Marketplace (ref=employer)
Female (ref=Male)
Age <35 (ref=age>35)
Hypertension (ref=checkup)
Office characteristics
FQHC
Family physician (ref=internal
medicine)
Solo practitioner (ref=4+ physicians)
2-3 practitioners (ref=4+ physicians)
County characteristics
Uninsured rate >25.5% (ref=13-25.5%)
Uninsured rate <13% (ref=13-25.5%)
Median income >$60k (ref=$45-$60k)
Median income <$45k (ref=$45-$60k)
State
Arkansas
Georgia
Iowa
Illinois
Massachusetts
Montana
New Jersey
Oregon
Pennsylvania
0.195
-0.154
0.102
0.035
-0.019
***
***
***
***
*
(0.164, 0.227)
(-0.173, -0.136)
(0.085, 0.119)
(0.015, 0.055)
(-0.036, -0.001)
-0.1
***
(-0.139, -0.062)
0.033
0.024
0.031
**
*
*
(0.012, 0.053)
(0, 0.047)
(0.002, 0.059)
0.059
0.011
0.022
-0.017
**
(0.025, 0.092)
(-0.02, 0.042)
(-0.001, 0.045)
(-0.042, 0.009)
0.028
-0.023
-0.005
0.021
-0.198
0.033
-0.04
-0.098
-0.046
***
***
*
(-0.013, 0.07)
(-0.062, 0.017)
(-0.049, 0.039)
(-0.028, 0.071)
(-0.25, -0.147)
(-0.03, 0.095)
(-0.083, 0.002)
(-0.142, -0.054)
(-0.091, 0)
Note: *P<.05,**P<.01, ***P<.001 Coefficients represent predicted margins from
regression models. Estimates from probit models can be interpreted as percentage
point differences. Standard errors are estimated using bootstrapping.
Source: Authors’ analysis of primary data from the 10 state primary care audit
study linked to data from 2014 data from SK&A and the 2014 County Health
Rankings.
4
Appendix Table 2. Models Reestimated withLogit Regression (as an Alternative to Probit)
Able to obtain a price
(N=7,865)
Estimate
95% CI
Caller characteristics
Uninsured (ref=employer)
Marketplace (ref=employer)
Female (ref=Male)
Age <35 (ref=age>35)
Hypertension (ref=checkup)
Office characteristics
FQHC
Family physician (ref=internal medicine)
Solo practitioner (ref=4+ physicians)
2-3 practitioners (ref=4+ physicians)
County characteristics
Uninsured rate >25.5% (ref=13-25.5%)
Uninsured rate <13% (ref=13-25.5%)
Median income >$60k (ref=$45-$60k)
Median income <$45k (ref=$45-$60k)
State
Arkansas
Georgia
Iowa
Massachusetts
Montana
New Jersey
Oregon
Pennsylvania
Texas
Referred to other sources
(N=3,396)
Estimate
95% CI
0.271
-0.146
0.082
0.012
-0.018
***
***
***
(0.24, 0.301)
(-0.168, -0.124)
(0.061, 0.104)
(-0.005, 0.029)
(-0.042, 0.006)
-0.417
-0.229
-0.27
0.037
-0.04
-0.185
0.028
0.075
0.045
***
*
***
**
(-0.235, -0.136)
(0.004, 0.052)
(0.041, 0.11)
(0.013, 0.076)
0.035
0.021
-0.007
-0.003
0.008
-0.024
-0.043
-0.033
-0.176
-0.015
-0.077
-0.028
-0.077
***
***
*
***
***
***
*
*
Told visit would be free
(N=4,469)
Estimate
95% CI
(-0.529, -0.306)
(-0.265, -0.193)
(-0.303, -0.238)
(0.007, 0.067)
(-0.076, -0.005)
-0.176
-0.054
-0.038
-0.022
-0.06
0.01
-0.029
-0.006
0.001
(-0.054, 0.074)
(-0.068, 0.01)
(-0.055, 0.042)
(-0.047, 0.049)
0.028
0.02
-0.011
0.005
(-0.023, 0.078)
(-0.006, 0.046)
(-0.041, 0.018)
(-0.022, 0.032)
(-0.002, 0.072)
(-0.023, 0.064)
(-0.039, 0.025)
(-0.034, 0.028)
0.009
-0.021
0.033
-0.067
(-0.05, 0.068)
(-0.074, 0.032)
(-0.012, 0.079)
(-0.12, -0.015)
0.015
-0.017
0.008
-0.06
(-0.024, 0.054)
(-0.061, 0.026)
(-0.02, 0.036)
*** (-0.09, -0.03)
(-0.047, 0.063)
(-0.079, 0.031)
(-0.101, 0.015)
(-0.099, 0.032)
(-0.246, -0.106)
(-0.102, 0.073)
(-0.119, -0.035)
(-0.08, 0.025)
(-0.135, -0.018)
-0.049
0.009
0.051
0.027
0.046
0.068
0.01
-0.058
0.045
(-0.128, 0.03)
(-0.079, 0.096)
(-0.041, 0.143)
(-0.067, 0.122)
(-0.054, 0.146)
(-0.049, 0.186)
(-0.072, 0.092)
(-0.157, 0.042)
(-0.047, 0.136)
-0.024
0.039
0.015
-0.025
0.075
-0.022
0.005
-0.004
0.028
*
***
***
***
**
***
*
**
(-0.196, -0.156)
(-0.079, -0.029)
(-0.055, -0.021)
(-0.038, -0.006)
(-0.079, -0.041)
(-0.07, 0.023)
(0.003, 0.075)
(-0.031, 0.061)
(-0.072, 0.021)
(0.027, 0.123)
(-0.088, 0.043)
(-0.031, 0.042)
(-0.056, 0.047)
(-0.02, 0.075)
5