fhircrackr: Flatten FHIR resources

  collapse = TRUE,
  comment = "#"

This vignette covers the following topics:

Before running any of the following code, you need to load the fhircrackr package:



In the vignette fhircrackr: Download FHIR resources you saw how to download FHIR resources into R. Now we'll have a look at how to flatten them into data.frames/data.tables. For the rest of the vignette, we'll work with the two example data sets from fhircrackr, which can be made accessible like this:

pat_bundles <- fhir_unserialize(bundles = patient_bundles)
med_bundles <- fhir_unserialize(bundles = medication_bundles)

See ?patient_bundles and ?medication_bundles for the FHIR search requests that generated them.

There are two extraction scenarios when you want to flatten FHIR bundles: Either you want to extract just one resource type, or you want to extract several resource types. Because the structure of different resource types is quite dissimilar, it makes sense to create one table per resource type. Therefore the result of the flattening process in fhircrackr can be either a single table (when extracting just one resource type) or a list of tables (when extracting more than one resource type). Both scenarios are realized with a call to fhir_crack(). We will now explain the two scenarios individually.

Extracting one resource type

We'll start with pat_bundles, which only contains Patient resources. To transform them into a table, we will use fhir_crack(). The most important argument fhir_crack() takes is bundles, an object of class fhir_bundle_list that is returned by fhir_search(). The second important argument is design, which tells the function which data to extract from the bundle. When we want to extract just one resource type, we can use a fhir_table_description in the argument design. fhir_crack() then returns a single data.frame or data.table (if argument data.tables = TRUE).

We'll show you an example of how it works first and then go on to explain the fhir_table_description in more detail.

pat_table_description <- fhir_table_description(
    resource = "Patient",
    cols     = list(
        id     = "id",
        gender = "gender",
        name   = "name/family",
        city   = "address/city"))

table <- fhir_crack(
    bundles = pat_bundles,
    design  = pat_table_description,
    verbose = 0)


The table_description

A fhir_table_description holds all the information fhir_crack() needs to create a table from resources of a certain type. It is created with fhir_table_description() and generally consists of the three elements that can be provided in the three following arguments:

The resource argument

This is basically a string that defines the resource type (e.g. Patient or Observation) to extract. You set it like this:

fhir_table_description(resource = "Patient")

Internally, fhir_table_description() calls fhir_resource_type() which checks the type you provided against list of all currently available resource types can be found at https://hl7.org/FHIR/resourcelist.html. Case errors are corrected automatically and the function throws a warning, if the resource type doesn't match the list under hl7.org.

As you can see in the above output, there are two more elements in a fhir_table_description which are filled automatically by fhir_table_description().

The cols argument

The cols argument takes the column names and XPath (1.0) expressions defining the columns to create from the FHIR resources. The XPath expression has to be built relatively to the root of the resource tree. If the cols element is empty, fhir_crack() will extract all available elements of the resource and name the columns automatically. To explicitly define columns, you can provide a (named) character or a (named) list with XPath expressions like this:

    resource = "Patient",
    cols     = list(
        gender = "gender",
        name   = "name/family",
        city   = "address/city"))

In this case a table with three columns called gender, name and city will be created. They will be filled with the value attribute that can be found under the respective xpath expressions in the resource. In the rare cases where you want to extract another attribute (besides value FHIR currently only allows id and url attributes), you have to add the attribute to the xpath expression like this: "node/node/@id". Note that this happens extremely rarely.

Internally, fhir_table_description() calls fhir_columns() to check the validity of the XPath expressions and assign column names. You can provide the XPath expressions in a named or unnamed character vector or a named or unnamed list. If you choose the unnamed version, the names will be set automatically and reflect the respective XPath expression:

#custom column names
fhir_columns(xpaths = c(
    gender = "gender",
    name   = "name/family",
    city   = "address/city"))

#automatic column names
fhir_columns(xpaths = c("gender", "name/family", "address/city"))

A fhir_columns object that is created explicitly like this can of course also be used in the columns argument of fhir_table_description. We strongly advise to only use fully specified relative XPath expressions here, e.g. "ingredient/strength/numerator/code" and not search paths like "//code", as those can generate unexpected results especially if the searched element appears on different levels of the resource.

The style argument

The final element of a fhir_table_description is a fhir_style object. This element controls how fhir_crack() deals with multiple entries to the same element and with columns that are completely empty, i.e. have only NA values. In the outputs above you can see, that the style takes some default values if you skip it in fhir_table_description(). You can change the defaults like this:

    resource = "Patient",
    cols     = list(
        gender = "gender",
        name   = "name/family",
        city   = "address/city"),
    style    = fhir_style(
        sep           = "||",
        brackets      = c("[", "]"),
        rm_empty_cols = FALSE))

As you can see, the style is created by a call to fhir_style() which can be used outside of fhir_table_description(), too:

fhir_style(sep = "||", brackets = c("[", "]"), rm_empty_cols = FALSE)

The sep element is a string defining the separator used when multiple entries to the same attribute are pasted together. This could for example happen if there is more than one address entry in a Patient resource. Examples of this are shown further down under the heading Multiple entries.

The brackets element is either an empty character (of length 0) or a character vector of length 2. If it is empty, multiple entries will be pasted together without indices. If it is of length 2, the two strings provided here are used as brackets for automatically generated indices to sort out multiple entries (see paragraph Multiple Entries). brackets = c("[", "]") e.g. will lead to indices like [1.1].

The rm_empty_cols: Can be TRUE or FALSE. If TRUE, columns containing only NA values will be removed, if FALSE, these columns will be kept.

All three elements of style can also be controlled directly by the fhir_crack() arguments sep, brackets and remove_empty_columns. If the function arguments are NULL (their default), the values provided in style are used, if they are not NULL, they will overwrite any values in style.

We will now work through examples using fhir_table_descriptions of different complexity.


Extract all available attributes

Lets start with an example where we only provide the (mandatory) resource component of the table_description. In this case, fhir_crack() will extract all available attributes and use default values for the style component:

#define a table_description
table_description1 <- fhir_table_description(resource = "Patient")

#convert resources
#pass table_description1 to the design argument
table <- fhir_crack(bundles = pat_bundles, design = table_description1, verbose = 0)

#have look at part of the results

As you can see, this can easily become a rather wide and sparse data frame. This is due to the fact that every attribute appearing in at least one of the resources will be turned into a variable (i.e. column), even if none of the other resources contain this attribute. For those resources, the value on that attribute will be set to NA. Depending on the variability of the resources, the resulting data frame can contain a lot of NA values. If a resource has multiple entries for an attribute (e.g. several addresses in a Patient resource), these entries will pasted together using the string provided in sep as a separator. The column names in this option are automatically generated by pasting together the path to the respective attribute, e.g. name.given.

Extract specific attributes

If we know which attributes we want to extract, we can specify them in a named list and provide it in the cols component of the data.frame description:

#define a table_description
table_description2 <- fhir_table_description(
    resource = "Patient",
    cols     = list(
        PID         = "id",
        use_name    = "name/use",
        given_name  = "name/given",
        family_name = "name/family",
        gender      = "gender",
        birthday    = "birthDate"))

#convert resources
#again pass table_description2 to the design argument
table <- fhir_crack(bundles = pat_bundles, design = table_description2, verbose = 0)

#have look at the results

This option will return more tidy and clear data frames, because you have full control over the extracted columns including their name in the resulting table. You should always extract the resource id, because this is used to link to other resources you might also extract.

If you are not sure which attributes are available or where they are located in the resource, it can be helpful to start by extracting all available attributes. If you are more comfortable with xml, you can also use xml2::xml_structure on one of the bundles from your bundle list, this will print the complete xml structure into your console. Then you can get an overview over the available attributes and their location and continue by doing a second, more targeted extraction to get your final data frame.

If you want to have a look at how the design looked that was actually used in the last call to fhir_crack() you can retrieve it with fhir_canonical_design().


As you can see the value attributes have been amended by fhir_crack(). This will be done automatically, whenever the XPath expressions don't state the attribute explicitly, so you don't have to mind that part.

Extracting more than one resource type

Of course the previous example is using just one resource type. If you are interested in several types of resources, you need one fhir_table_description per resource type. You can bundle a bunch of fhir_table_descriptions in a fhir_design. This is basically a named list of fhir_table_descriptions, and when you pass it to fhir_crack(), the result will be a named list of tables with the same names as the design. Consider an example where we have downloaded MedicationStatements referring to a certain medication as well as the Patient resources these MedicationStatements are linked to.

The design to extract both resource types could look like this:

meds <- fhir_table_description(
    resource = "MedicationStatement",
    cols     = list(
        ms_id       = "id",
        status_text = "text/status",
        status      = "status",
        med_system  = "medicationCodeableConcept/coding/system",
        med_code    = "medicationCodeableConcept/coding/code",
        med_display = "medicationCodeableConcept/coding/display",
        dosage      = "dosage/text",
        patient     = "subject/reference",
        last_update = "meta/lastUpdated"),
    style    = fhir_style(
        sep           = "|",
        brackets      = NULL, 
        rm_empty_cols = FALSE))

pat <- fhir_table_description(resource = "Patient")

design <- fhir_design(meds, pat)

In this example, we have spelled out the table_description MedicationStatement completely, while we have used a short form for Patients. It looks like this:


As you can see, each table_description is identified by a name, which will also be the name of the corresponding table in the result of fhir_crack().

You can assign the names explicitly, if you prefer:

design <- fhir_design(Medications = meds, Patients = pat)


And you can also extract single table_descriptions by their name:


We can use the design for fhir_crack():

list_of_tables <- fhir_crack(bundles = med_bundles, design = design, verbose = 0)



As you can see, the result is a list of two data frames, one for Patient resources and one for MedicationStatement resources. When you use fhir_crack() with a fhir_desgn() instead of a fhir_table_description, the result is an object of class fhir_df_list or fhir_dt_list that also has the design attached. You can extract the design from a list like this using fhir_design():


Note that this doesn't work on single tables created with a fhir_table_description.

Saving and reading designs

If you want to save a design for later or to share with others, you can do so using the fhir_save_design(). This function takes a design and saves it as an xml file:

temp_dir <- tempdir()
fhir_save_design(design = design, file = paste0(temp_dir, "/design.xml"))

To read the design back into R, you can use fhir_load_design():

fhir_load_design(paste0(temp_dir, "/design.xml"))

Multiple entries

A particularly complicated problem in flattening FHIR resources is caused by the fact that there can be multiple entries to an attribute. The profile according to which your FHIR resources have been built defines how often a particular attribute can appear in a resource. This is called the cardinality of the attribute. For example the Patient resource defined here can have zero or one birthdates but arbitrarily many addresses. In general, fhir_crack() will paste multiple entries for the same attribute together in the data frame, using the separator provided by the sep argument. In most cases this will work just fine, but there are some special cases that require a little more attention.

Let's have a look at an example bundle containing just three Patient resources. You can make it available in your workspace like this:

bundle <- fhir_unserialize(example_bundles2)

This is how the xml looks:

        <id value='id1'/>
            <use value='home'/>
            <city value='Amsterdam'/>
            <type value='physical'/>
            <country value='Netherlands'/>
            <given value='Marie'/>
        <id value='id2'/>
            <use value='home'/>
            <city value='Rome'/>
            <type value='physical'/>
            <country value='Italy'/>
            <use value='work'/>
            <city value='Stockholm'/>
            <type value='postal'/>
            <country value='Sweden'/>
            <given value='Susie'/>

        <id value='id3'/>
            <use value='home'/>
            <city value='Berlin'/>
            <type value='postal'/>
            <country value='France'/>
            <use value='work'/>
            <city value='London'/>
            <type value='postal'/>
            <country value='England'/>
            <given value='Frank'/>
            <given value='Max'/>


This bundle contains three Patient resources. The first resource has just one entry for the address attribute. The second Patient resource has two entries containing the same elements for the address attribute. The third Patient resource has a rather messy address attribute, with three entries containing different elements and also two entries for the name attribute.

Let's see what happens if we extract all attributes:

desc1 <- fhir_table_description(resource = "Patient", style = fhir_style(sep = " | "))

df1 <- fhir_crack(bundles = bundle, design = desc1, verbose = 0)


As you can see, multiple entries for the same attribute (address and name) are pasted together. This works fine for Patient 2, but for Patient 3 you can see a problem with the number of entries that are displayed. The original Patient resource had three (incomplete) address entries, but because the first two of them use complementary elements (use and city vs. type and country), the resulting pasted entries look like there had just been two entries for the address attribute.

You can counter this problem by setting brackets:

desc2 <- fhir_table_description(
    resource = "Patient",
    style    = fhir_style(
        sep      = " | ",
        brackets = c("[", "]")))

df2 <- fhir_crack(bundles = bundle, design = desc2, verbose = 0)


Now the indices display the entry the value belongs to. That way you can see that Patient resource 3 had three entries for the attribute address and you can also see which attributes belong to which entry.

Of course the above example is a very specific case that only occurs if your resources have multiple entries with complementary elements. In the majority of cases multiple entries in one resource will have the same structure, thus making numbering of those entries superfluous. But the indices also help to disentangle those entries and put them in separate rows, as you'll see in the next paragraph.

Process Data Frames with multiple Entries

Melt data frames with multiple entries

If the data frame produced by fhir_crack() contains multiple entries, you'll probably want to divide these entries into distinct observations at some point. This is where fhir_melt() comes into play. fhir_melt() takes an indexed data frame with multiple entries in one or several columns and spreads (aka melts) these entries over several rows:

r fhir_melt( indexed_data_frame = df2, columns = "address.city", brackets = c("[", "]"), sep = " | ", all_columns = FALSE)

The new variable resource_identifier maps which rows in the created data frame belong to which row (usually equivalent to one resource) in the original data frame. brackets and sep should be given the same character vectors that have been used to build the indices in fhir_melt(). columns is a character vector with the names of the variables you want to melt. You can provide more than one column here but it makes sense to only have variables from the same repeating attribute together in one call to fhir_melt():

```r cols <- c("address.city", "address.use", "address.type", "address.country")

fhir_melt( indexed_data_frame = df2, columns = cols, brackets = c("[", "]"), sep = " | ", all_columns = FALSE) `` If the names of the variables in your data frame have been generated automatically withfhir_crack()you can find all variable names belonging to the same attribute withfhir_common_columns()`:

r cols <- fhir_common_columns(data_frame = df2, column_names_prefix = "address") cols With the argument all_columns you can control whether the resulting data frame contains only the molten columns or all columns of the original data frame:

r fhir_melt( indexed_data_frame = df2, columns = cols, brackets = c("[", "]"), sep = " | ", all_columns = TRUE) Values on the other variables will just repeat in the newly created rows.

If you try to melt several variables that don't belong to the same attribute in one call to fhir_melt(), this will cause problems, because the different attributes won't be combined correctly:

```r cols <- c(cols, "id") fhir_melt( indexed_data_frame = df2, columns = cols, brackets = c("[", "]"), sep = " | ", all_columns = TRUE)


Instead, melt the attributes one after another:

cols <- fhir_common_columns(data_frame = df2, column_names_prefix = "address")

molten_1 <- fhir_melt(
    indexed_data_frame = df2,
    columns            = cols,
    brackets           = c("[", "]"),
    sep                = " | ",
    all_columns        = TRUE)


molten_2 <- fhir_melt(
    indexed_data_frame = molten_1,
    columns            = "name.given",
    brackets           = c("[", "]"),
    sep                = " | ",
    all_columns        = TRUE)


This will give you the appropriate product of all multiple entries.

Remove indices

Once you have sorted out the multiple entries, you might want to get rid of the indices in your data.frame. This can be achieved using fhir_rm_indices():

fhir_rm_indices(indexed_data_frame = molten_2, brackets = c("[", "]"))

Again, brackets and sep should be given the same character vector that was used for fhir_crack() and fhir_melt() respectively.

file.remove(paste0(temp_dir, "/design.xml"))

Try the fhircrackr package in your browser

Any scripts or data that you put into this service are public.

fhircrackr documentation built on June 17, 2021, 5:12 p.m.