Overview of Provider • provider

library(provider)
library(dplyr)
library(purrr)

The overarching goal of provider is to make the experience of accessing publicly-available Provider data easier and more consistent across a variety of sources. It aims to accomplish this through the following goals:

Non-Destructive Tidy Output, with an option to turn it off by setting a function’s tidy parameter to FALSE.
Variable Standardization, for the express purpose of making it easier to understand (and make connections between) each API’s output. This will also allow for the removal of duplicate information and greatly simplify the process of merging data across outputs.
Input Validation. Not only is this simply good practice, it also prevents unnecessary querying of APIs.
Helpful Documentation. The impetus for this package was to cobble together a motivating example of using a programming language to streamline the process of data acquisition for, among other things, medical coding, billing, and healthcare revenue cycle management. Though it’s now grown beyond that, the intended audience remiains the same: non-programmers in healthcare who are interested in what a programming language like R can do to make their work easier. As such, the documentation is written with in a way that assumes no prior knowledge of R or programming in general. Domain-specific terminology and concepts are explained in detail, as there is not one person in existence that understands every aspect of the business of health care, including the author of this package.

Tidy Output

In tidy data, every column is a variable, every row is an observation, and every cell is a single value.

Many of the outputs of these APIs violate this form by

Storing column headers as values, not variable names
Storing multiple variables in one column

To remedy this, there is some post-processing done to the output before it is returned. Each function has a tidy parameter that is set to TRUE by default:

order_refer(npi = 1043477615)
#> # A tibble: 5 × 4
#>   npi        first last    eligible                 
#>   <chr>      <chr> <chr>   <fct>                    
#> 1 1043477615 SARAH HUSSAIN Medicare Part B          
#> 2 1043477615 SARAH HUSSAIN Home Health Agency       
#> 3 1043477615 SARAH HUSSAIN Durable Medical Equipment
#> 4 1043477615 SARAH HUSSAIN Power Mobility Devices   
#> 5 1043477615 SARAH HUSSAIN Hospice

Setting tidy to FALSE will return a data.frame of the API’s un-processed output:

order_refer(npi = 1043477615, tidy = FALSE)
#>          NPI LAST_NAME FIRST_NAME PARTB DME HHA PMD HOSPICE
#> 1 1043477615   HUSSAIN      SARAH     Y   Y   Y   Y       Y

Several functions also have arguments that implement additional optional transformations to the output, such as pivot:

order_refer(npi = 1043477615, pivot = FALSE)
#> # A tibble: 1 × 8
#>   npi        last_name first_name partb dme   hha   pmd   hospice
#>   <chr>      <chr>     <chr>      <lgl> <lgl> <lgl> <lgl> <lgl>  
#> 1 1043477615 HUSSAIN   SARAH      TRUE  TRUE  TRUE  TRUE  TRUE

The tidy transformations applied to the output of each function include:

Favoring a long format over wide.
More descriptive, thematic variable names involving prefixes and a standardized vocabulary across all outputs
Various quality-of-life data transformations (e.g., using snake_case for variable names, replacing empty character cells with NA; converting years, dates, booleans to their respective data types)
Nesting large groups of related columns into lists, to initially emphasize the most important output.

However, in the event that you would prefer to do your own transformations, simply turn it off with tidy = FALSE.

Helper Functions

Several functions have a required year argument. The years available to query these APIs might change at any time, so there is an accompanying helper function to retrieve the years available, in the form of <function_name>_years(). For instance, quality_payment()’s is:

qpp_years()
#> [1] 2017 2018 2019 2020 2021 2022

These can also be used in a pipeline, for searching all available years:

map_dfr(qpp_years(), ~quality_payment(year = .x, npi = 1043477615)) |> 
  select(year, 
         participation_type, 
         beneficiaries,
         services,
         charges,
         final_score,
         pay_adjust, 
         org_size, 
         apms_entity_name)
#> Error in `map()`:
#> ℹ In index: 6.
#> Caused by error in `dplyr::arrange()`:
#> ℹ In argument: `..2 = participation_type`.
#> Caused by error:
#> ! object 'participation_type' not found

Several functions have parallelized versions, denoted by an underscore at the end of their name (_):

quality_payment_(npi = 1043477615) |> 
  select(year, 
         participation_type, 
         beneficiaries,
         services,
         charges,
         final_score,
         pay_adjust, 
         org_size, 
         apms_entity_name)
#> Error:
#> ℹ In index: 6.
#> Caused by error in `dplyr::arrange()`:
#> ℹ In argument: `..2 = participation_type`.
#> Caused by error:
#> ! object 'participation_type' not found

Validation Checks

Several validation checks have been implemented, including checks for NPIs, PAC IDs, and Enrollment IDs:

National Provider Identifier: A National Provider Identifier (NPI) is a unique 10-digit identification number issued to health care providers in the United States by the Centers for Medicare and Medicaid Services (CMS) through the National Plan and Provider Enumeration System (NPPES). All individual HIPAA–covered healthcare providers or organizations must obtain an NPI. Once assigned, a provider’s NPI is permanent and remains with the provider regardless of job or location changes.

# Must be 10 digits long
open_payments(year = 2021, npi = 12345691234)
#> Error in `open_payments()`:
#> ! An NPI must be 10 digits long.
#> ✖ 12345691234 contains 11 digits.

# Must be numeric
nppes(npi = "O12345678912")
#> Error in `nppes()`:
#> ! An NPI must be numeric.
#> ✖ "O12345678912" contains non-numeric characters.

# Must pass Luhn check
pending(npi = 001234569123, type = "P")
#> Error in `pending()`:
#> ! "1234569123" is not a valid NPI.
#> → Did you mean "1234569121"?

Provider Associate-level Control ID: A Provider associate-level control ID (PAC ID) is a 10-digit unique numeric identifier that is assigned to each individual or organization in PECOS. The PAC ID links all entity-level information (e.g., tax identification numbers and organizational names) and may be associated with multiple enrollment IDs if the individual or organization enrolled multiple times under different circumstances.

# Must be 10 digits long
affiliations(pac = 0123456789)
#> Error in `affiliations()`:
#> ! A PAC ID must be 10 digits long.
#> ✖ 123456789 contains 9 digits.

# Must be numeric
hospitals(pac_org = "O12345678912")
#> Error in `hospitals()`:
#> ! A PAC ID must be numeric.
#> ✖ "O12345678912" contains non-numeric characters.

Medicare Enrollment ID: An Enrollment ID is a 15-digit unique alphanumeric identifier that is assigned to each new provider enrollment application. All enrollment-level information (e.g., enrollment type, enrollment state, provider specialty and reassignment of benefits) is linked through the Enrollment ID.

# Must be a character vector
clinicians(enid = 0123456789123456)
#> Error in `clinicians()`:
#> ! An Enrollment ID must be a <character> vector.
#> ✖ 123456789123456 is a <numeric> vector.

# Must be 15 characters long
reassignments(enid = "I123456789123456")
#> Error in `reassignments()`:
#> ! An Enrollment ID must be 15 characters long.
#> ✖ "I123456789123456" contains 16 characters.

# Must begin with a capital I (Individual) or O (Organization/Group)
providers(enid = "L12345678912345")
#> Error in `providers()`:
#> ! An Enrollment ID must begin with a capital `I` or `O`.
#> ✖ "L12345678912345" begins with "L".

# Some functions require one of ID types
hospitals(enid_org = "I20180115000174")
#> Error in `hospitals()`:
#> ! An Organizational Enrollment ID must begin with a capital `O`.
#> ✖ "I20180115000174" begins with "I".