Data on the names, credentials and addresses of the providers in the map are derived from the National Plan & Provider Enumeration System (NPPES). CMS developed the NPPES to assign unique identifiers, known as National Provider Identifiers (NPIs), to health care providers. The health care provider’s address information is collected at the time of enrollment and updated periodically. Note that physician addresses are not continuously updated, nor or they current (i.e., they reflect data as of 2018).

Physicians’ addresses were cleaned and standardized prior to geocoding (mapped to longitude and latitude) using Google Maps Geocoding Application Programming Interface (API) and US Census Bureau API. Data cleaning and geocoding was processed using R version 4.0.2 and associated packages (www.r-project.org). The interactive physician map was created using Python version 3.7 and associated packages.

Data used to create the map come from the 2018 Physician and Other Supplier PUF, which is generated from CMS administrative claims data for Medicare beneficiaries enrolled in the fee-for-service (FFS) program. Weighted mortality risk scores were generated using 100% of Medicare enrollment and fee-for-service claims data summarized from the Master Beneficiary Summary File (MBSF) and Part B non-institutional claims (NCH Claim Type Codes '71' and '72').

Model coefficients of the hazard ratio for COVID-19 death were derived from a validated COVID-19 mortality risk calculator (Jin, J., Agarwala, N., Kundu, P. et al., 2020). Coefficients from the model found in Supplementary Table 1 were applied to calculate risk for mortality for beneficiaries 65 and older. Note, we did not make any adjustments for geography. We excluded beneficiaries younger than 65 since eligibility for Medicare under 65 represents a unique cohort with a more complex profile than the under-65 population represented in the general model. The model developed by Jin, J., Agarwala, N., Kundu, P. et al. used information on the prevalence and co-occurrence of risk factors from various national databases to make population-level projections of risk associated with predisposing factors for the adult population across 477 US cities and for the 65-years-and-older population enrolled in Medicare across 3,113 US counties. Not all factors from the risk model were scored using the Medicare data given the unavailability of certain metrics in claims data. Factors included in our risk calculations and their respective coefficients are included in Supplemental Figure 1. Supplemental Figure 2 shows the distribution of the risk ratio’s for beneficiaries scored and is consistent with the distribution of scores in the original model.

Supplemental Figure 1. Model coefficient for each category of various risk factors.

1. Chronic and acute conditions were calculated using ICD-10 diagnoses from 2018 Part B non-institutional claims and categorized using Chronic Conditions Warehouse definitions unless otherwise specified.
2. Respiratory disease used all ICD-10 “J” codes, with the exception of asthma.
3. Uncontrolled diabetes was differentiated from controlled diabetes if the beneficiary had a claim for a hospitalization with a diabetes DRG.
4. Cancers within the previous year were flagged where there was a claim for a hospitalization with a cancer DRG (e.g., treatment); a diagnosis of cancer was not used to score the model; treatment or history of cancer >1 year was not scored.
5. BMI is unavailable in claims data; a proxy for obesity was used if CPT codes 99401 or 99402 were submitted in a claim and scored as Class II Obesity

Supplemental Figure 2. Distribution of COVID-19 Risk Scores for Medicare Beneficiaries Age 65+

Supplemental Figure 3 shows the proportion of the aged population in each of the risk categories for the research undertaken by Jin, J., Agarwala, N., Kundu, P. et al., derived from the Centers for Disease Control and Prevention (CDC) National Health Interview Survey (NHIS), and of that derived from the Medicare FFS provider data represented in the map.

Supplemental Figure3. COVID-19 Risk Categories for People Age 65+

Limitations

While the Physician and Other Supplier PUF has a wealth of information on payment and utilization for Medicare Part B services, the dataset has a number of limitations. Of particular importance is the fact that the data may not be representative of a physician’s entire practice as it only includes information on Medicare FFS beneficiaries. Additionally, CMS suppression policy dictates that no data be reported for a cell size of fewer than 11 beneficiaries. The result of data suppression impacts providers who see a fewer number of Medicare FFS beneficiaries and may disproportionately affect rural and/or providers with smaller practices. Second, risk score metrics are only calculated for the population of Medicare FFS beneficiaries who are 65+, thus limiting the sample of scored beneficiaries; as a result, not all mapped providers may have an average COVID-19 risk score.

Resources

Jin, J., Agarwala, N., Kundu, P. et al. Individual and community-level risk for COVID-19 mortality in the United States. Nat Med (2020). https://doi.org/10.1038/s41591-020-01191-8

See: https://www.nature.com/articles/s41591-020-01191-8#Sec2

See: https://covid19risktools.com:8443/

Medicare Provider Utilization and Payment Data: Physician and Other Supplier Public Use Files

See: https://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports/Medicare-Provider-Charge-Data/Physician-and-Other-Supplier

Building Confidence in COVID-19 Vaccines Among Your Patients Tips for the Healthcare Team

See: https://www.cdc.gov/vaccines/covid-19/downloads/VaccinateWConfidence-TipsForHCTeams_508.pdf

Google Maps APIs. See: https://developers.google.com/maps/documentation/geocoding/start

US Census Bureau APIs. See: https://geocoding.geo.census.gov/