Differences in Primary Care Appointment Availability and Wait
Times by Neighborhood Characteristics: a Mystery Shopper Study
David Grande, MD, MPA
, Jessica X. Zuo, BA
, Rathnam Venkat, BA
, Xinwei Chen, MS
Katelyn R. Ward, BA
, and Nandita Mitra, PhD
Division of General Internal Medicine, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA;
Leonard Davis Institute of
Health Economics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA;
Perelman School of Medicine, University of
Pennsylvania, Philadelphia, PA, USA;
Wharton School, University of Pennsylvania, Philadelphia, PA, USA;
Tufts University School of Medicine, Boston,
Department of Epidemiology, Boston University School of Public Health, Boston, MA, USA;
Department of Biostatistics and Epidemiology,
Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA.
KEY WORDS: primary care access; simulated patient; primary care supply;
appointment availability; appointment wait times.
J Gen Intern Med
© Society of General Internal Medicine 2018
Primary care is widely recognized as a gateway to the
health care system and improved health.
have found disparities in appointment access by insurance
(commercial vs. Medicaid) and mixed findings on
the association of neighborhood socioeconomic status and
having a usual source of care.
We previously found
large racial differences in the supply of primary care
across neighborhoods in a large urban area (Philadelphia,
In this study, we examine how appointment access
varies by neighborhood socio-demographics and primary
care supply—hypothesizing less access in low SES neigh-
borhoods and those with lower supply.
As previously described, we inventoried adult primary care
practices in and near Philadelphia County.
From July 6 to
September 14, 2015, research assistants posing as patients
called practices to request a new, non-urgent, appointment.
Medicaid participating practices received private and
Medicaid-insured calls. Practices were excluded if they did
not offer primary care (n = 16), had a disconnected phone (n =
17), or served a specialized population (n = 16) (e.g., univer-
sity student health center). The University of Pennsylvania
Institutional Review Board approved this study.
We used the American Community Survey (2008–2013) to
determine characteristics of census tracts and tract-level
population-to-provider ratios for adult primary care.
a relative measure of low primary care access
; clusters of five
or more contiguous census tracts in the lowest quintile for
primary care supply.
We defined availability and wait times as whether an appoint-
ment was offered (binary) and the number of days from the
request to the appointment offered (continuous), respectively.
We modeled the association of availability with census tract
characteristics and our measure of low primary care supply
using a multi-level random intercept logistic mixed effects
model with clustering at the clinic and census tract levels. A
linear mixed effects model was similarly used to examine wait
We identified 399 practices in the study area— 276 accepted
Medicaid and private insurance, 111 accepted only private
insurance, and 12 were only reached by our Medicaid callers.
We excluded 12 due to inability to complete scheduling.
Callers were offered an appointment in 79% of the calls;
fewer Medicaid callers received appointments than those
with private insurance (68 vs. 87%; OR = 0.32; 95%
CI = 0.21–0.47). There were no differences in appoint-
ment availability for privately insured callers at Medicaid
participating vs. non-participating practices. Practices’
census tract characteristics were not associated with ap-
pointment availability except for the uninsured rate (Ta-
ble 1). Offices in census tracts with high rates (≥ 30%) of
uninsurance were less likely to offer appointments (OR =
0.24; 95% CI = 0.07–0.90). There was no association
with location in a lower primary care supply area.
Appointment Wait Time
The median appointment wait time was 10 days (IQR = 4–21).
Wait times were similar (p = 0.29) for Medicaid-insured (me-
dian = 12.5, IQR = 5–25.5) and privately insured callers (me-
dian=9,IQR=4–20). Findings were similar when the sample