Appointment wait time data for primary & specialty care in veterans health administration facilities vs. community medical centers

The datasets summarized in this article include more than 38 million appointment wait times that U.S. military veterans experienced when seeking medical care since January 2014. Our data include both within Veterans Health Administration (VHA) facilities and community medical centers, and wait times are stratified by primary/specialty care type. Deidentified wait time data are reported at the referral-level, at the VHA facility-level, and at the patient's 3-digit ZIP code-level. As of this writing, no other U.S. health care system has made their wait times publicly available. Our data thus represent the largest, national, and most representative measures of timely access to care for patients of both VHA and community providers. Researchers may use these datasets to identify variations in appointment wait times both longitudinally and cross-sectionally, conduct research on policies and interventions to improve access to care, and to incorporate fine-grained measures of wait times into their analyses.


a b s t r a c t
The datasets summarized in this article include more than 38 million appointment wait times that U.S. military veterans experienced when seeking medical care since January 2014. Our data include both within Veterans Health Administration (VHA) facilities and community medical centers, and wait times are stratified by primary/specialty care type. Deidentified wait time data are reported at the referral-level, at the VHA facility-level, and at the patient's 3-digit ZIP codelevel. As of this writing, no other U.S. health care system has made their wait times publicly available. Our data thus represent the largest, national, and most representative measures of timely access to care for patients of both VHA and community providers. Researchers may use these datasets to identify variations in appointment wait times both longitudinally and cross-sectionally, conduct research on policies and inter-ventions to improve access to care, and to incorporate finegrained measures of wait times into their analyses.
© 2021 The Author(s

Value of the Data
• There are currently not nationwide, publicly available datasets of appointment wait times within the United States. • Our data provide a unique opportunity for researchers and data journalists to measure wait times for veterans to access care both within the VHA and in the community, both crosssectionally and over time. • Facility, county, and referral-level data describe substantial variation in appointment wait times for VHA and community-based providers across a broad range of specialties. • The large sample size, nationwide coverage, consistent data collection, and broad range of appointment types provide several advantages over previously-published estimates of wait times. • Researchers may leverage these and other datasets to study the relationship between health policies, appointment wait times, and a wide variety of health, economic, and social outcomes.

Data Description
Prior to 2014, the Veterans Health Administration (VHA) only reimbursed providers in the community who provided medical care to veterans when the VHA was unable to do so (e.g. nearby facilities did not have certain types of specialists) or for emergency care [1] . The 2014 Veteran Health Administration (VHA) wait time scandal prompted a nationwide investigation into the amount of time Veterans spent waiting to receive care, and whether their delayed access contributed to significant adverse health outcomes [2] . Congress responded by passing the Veteran's Access to Care through Choice, Accountability, and Transparency Act of 2014, which authorized $16 billion for the Veterans Choice Program (VCP). Under the VCP, veterans who live more than 40 miles from the nearest VHA facility or could not schedule an appointment within 30 days were now permitted to receive care through community providers who contract with the VHA [3] . Congress expanded VCP eligibility criteria in 2015 to include Veterans with an "unusual and excessive burden for travel to VHA health care facilities," such as geographic challenges, medical conditions, and environmental conditions like road blockages and traffic [ 4 , 5 ]. The MISSION Act of 2018 further expanded Veterans' eligibility to access community care options and included additional interventions focused on telehealth and mobile deployment units to expand avenues for Veterans to interact with the health care system [6] . Eligible veterans may now seek VHA-funded care from community providers if their estimated drive time to the nearest VHA facility exceeds 60 min, replacing the VCP's 40 mile eligibility standard.
The VHA Corporate Data Warehouse (CDW) contains a record for every referral to primary or specialty care, regardless of whether patients are seen at a VHA facility or community medical center. We observe dates for when referrals were requested, dates when appointments were schedule, and dates when appointments were completed. A consult status of "completed" indicates an initial encounter the healthcare provider who received the referral; additional followup appointments and procedures may occur after this date. We also observe primary/specialty type for each consult. Note that the VHA uses "stop codes" to identify care type; stop codes are 3-digit identifiers used to identify the work group primarily responsible for providing a clinical service, and are used for purposes of workload credit, managerial accounting, and program evaluation (see Table 1 for a list) [7] . These stop codes are unique to the VHA but have been grouped together by researchers to study primary care [8] , mental health [9] , and other specialties [ 10 , 11 ]. Local VHA facilities must first approve all referrals to community providers; we also observe dates of approval for these requests. Additional details on the consult request process are outlined in VHA Direction 1232(2) [12] .
Our data source thus incorporates the universe of primary and specialty care appointments paid for by the VHA from January 2014 through April 2021. The associated Mendeley data repository will be updated approximately quarterly with new data as they become available.
We calculated three types of appointment wait times by specialty: (1) Consult-level wait times wait times which include specialty type, year, whether a VHA or community provider were used, wait times, and patient's 3-digit ZIP Code. (2) County-level wait times which aggregates all appointment requests by patient's county of residence. (3) Facility-level wait times which aggregates all appointment requests to the VHA parent facility which provided approval. A parent facility is referred to as a "station" or "STA3N" within the VHA and may also have several subsidiary medical centers or community-based outpatient clinics assigned to it.
These datasets cover 41,249,208 consult requests for both primary and specialty care during the time period from January 1, 2014 through December 31, 2020. We fill an important data gap in U.S. health services research, which until now has lacked a large national dataset on appointment wait times for either primary or specialty care. We provide researchers and journalists with the broadest, most rigorously-collected datasets on wait times that are publicly-available. Data dictionaries for each dataset are available in Tables 2-4 .

Experimental Design, Materials and Methods
We used SQL to query the VHA CDW and calculate wait times for referrals to both VHA and community-based providers. Referrals with completed, discontinued, or canceled status were included for calculations. Discontinued & cancelled appointments accounted for 2.3% and 1.7% of total consult volume respectively, and were included since their exclusion may bias estimates of wait times downwards (e.g. if a Veteran is unsatisfied with the wait and thus cancels their appointment). Referrals were excluded if an appointment was never scheduled, since no wait time was observed. Referrals were also excluded if they were missing information on facility or primary/specialty care type. Note the terms 'consults' and 'referrals' are used interchangeably within the VHA.
The CDW's Con.Consult table identifies the facility where the consult was created, a unique patient identifier, initial request date, and may be linked to other tables to identify consult type (e.g. cardiology, gastroenterology). The Con.ConsultActivity table tracks changes to the status of a consult and contains individual rows for when a consult is created, approved, scheduled, completed, cancelled, or discontinued. We use the 'ActivityDateTime' field to calculate four outcome measures: (1) Days to Approved, a measure of the difference between dates for when a consult is created and when it has been approved by the local VHA medical center. For community care, this is when the veteran was authorized to seek care in the community. A violin plot of approval wait times for four high-volume medical specialties is contained in Fig. 1 . A violin plot is similar to a box plot with the addition of a rotated kernel density plot on each side which shows the distribution of the data. (2) Days to Scheduled, a measure of the difference between when a consult is approved and when the appointment is scheduled. For community care, this measure represents the date the local VHA medical center followed up with a Veteran and found out they have scheduled the appointment; this is likely several days or weeks after the Veteran actually made the appointment. (3) Days to Completed, a measure of the difference between when a consult is approved and when it was completed. (4) Total Wait Time, a measure of the difference between when a referral was initially requested and when the appointment was completed. For cancelled/discontinued appointments, this is the difference between when a referral was initially requested and the scheduled appointment date. A scatter plot of wait times for VHA and community care at the ZIP-3 is displayed in Fig. 2 . On average, the VHA outperformed community medical centers in terms of mean wait times. Further, VHA wait times were positively correlated with wait times at community medical centers.
The consult tables were also linked to the Appt.Appointments table through a unique Con-sultSID, which allows us to observe actual appointment dates. These appointment dates were validated by chart reviews. We leveraged the ToRequestServiceName field of the Con.Consult CDW table to identify and exclude consultation types that had average completion times of < = 0.2 days. Chart reviews indicated these are mostly e-consultations (such as email or text messages between providers) that are opened and closed within a few minutes or hours.
Our referral-level wait time dataset indicates appointment year, wait time measures, 3-digit ZIP Code of the veteran's home address (obtained from the SPatient.SPatientAddress table), an indicator for whether the appointment was for a VHA or community provider, and the primary stop code. VHA uses primary stop codes (also known as Decision Support System Identifiers) to identify the main clinical group responsible for a patient's care (see Table 4 ). We created a facility-level dataset by averaging appointment wait times by each stop code in a given month.
The resulting referral-level dataset was then aggregated to calculate mean average wait times by month at the ZIP code-and VHA facility-level, then deidentified for public release. All data preparation was performed in Microsoft SQL Server Management Studio version 15.10.18206.0 (Redmond, WA). The latest SQL script used to calculate the three wait time datasets, as well as copies of each dataset, are publicly available within our Mendeley Data repository.
We note several important caveats with these data. Prior to 2018, there was no standardized method for VHA facilities to indicate whether or not a referral was to VHA or community-based providers. We identified referrals to the community by text searches of the 'ToRequestService-Name' field of the Con.Consult CDW table (e.g. mentions of 'community care,' 'CHOICE,' 'fee basis'). We estimate that approximately 50% to 75% of community-based consultations were misclassified as VHA consultations before May 2018. The number of non-VA consults that we can identify increased sharply starting in 2018 ( Fig. 3 ). This comports with guidance which went out on how to record these consults in the data (e.g. use of stop code 669 and including the phrase 'COMMUNITY CARE' in the 'ToRequestServiceName' field of the Con.Consult CDW table). The implementation of stop code '669' has enabled better identification of community care consults. Unfortunately, this general stop code has also made it more difficult to identify their specialty. We follow a tiered approach to try and convert these 669 stop codes; in our tests, 87% of stop codes are matched to more informative stop codes.
Lastly, VHA users who would like to run our code are advised not to examine wait times within the previous six months. Appointment information, especially for community care consults, may only appear in the CDW after long and variable lags of several months.

File inventory
• Wait time data at the facility level (processed).
• Wait time data at the county level (processed).
• Wait time data at the consultation level (processed).
• SQL script to calculate wait time datasets.

Ethics Statement
The Privacy Office of the Veterans Affairs Boston Healthcare System have certified these datasets are de-identified and may be publicly-released as part of this publication.

Declaration of Competing Interest
Yevgeniy Feyman, Aaron Legler, and Kevin Griffith are investigators at the VA Boston Healthcare System. The content is solely the responsibility of the authors and does not necessarily represent the views of the VHA, which did not have editorial input or control over this research.