def vuid2rxcui():
"""
Creates a .csv file containing the results of the RxNorm API
Allows calls to be made only once & can be merged onto data cube
Input: aggregated source data as a dataframe
Output: CSV file with all unique VUIDs and corresponding RxNorm CUIs
"""
if os.path.exists('./vuid_to_rxcui.csv'):
if config.print_status:
print('VUIDs already created in .csv file')
return None
# load data & find unique values
df = load_agg_data()
vuids = df.VUID.unique()
if config.print_status:
print('Initiating API requests for VUIDs')
# make iterable list of URLs to send to api
urls = api_rxnorm.make_url_requests(vuids, 'vuid')
results = api_rxnorm.async_calls(urls)
# clean up & store in CSV
vuid_to_rxcui = defaultdict(list)
for result in results:
# extract digits numeric results
digits = re.findall(b'\d+', result)
vuid = int(digits[0])
# could be multiple rxcuis for a single vuid
cuis = []
for d in digits[1:]:
cuis.append(int(d))
vuid_to_rxcui[vuid] = cuis
vuid_to_rxcui = pd.DataFrame.from_dict(vuid_to_rxcui, orient='index')
vuid_to_rxcui.reset_index(inplace=True)
assert (len(vuid_to_rxcui) == len(vuids) - 1)
# make enough column names for all extra placeholders
cnames = ['vuid']
for i in range(1, vuid_to_rxcui.shape[1]):
cnames.append('rxcui' + str(i))
vuid_to_rxcui.columns = cnames
if config.print_status:
print('API requests for VUID to RxCUI Complete.')
print('Initiating API requests for generic & brand RxCUIs...')
# search for generic & brand names
rxcui_brand_generic = api_rxnorm.getBrandGeneric(vuid_to_rxcui,
cui_col='rxcui1')
# coerce from dictionary to dataframe to allow merge
df = pd.DataFrame.from_dict(rxcui_brand_generic, orient='index')
df.rename(columns={0: 'rxcui1_brand', 1: 'rxcui1_generic'}, inplace=True)
df['rxcui1'] = df.index.astype(float)
# merge & save
vuid_to_rxcui = vuid_to_rxcui.merge(df, how='left', on='rxcui1')
vuid_to_rxcui.to_csv('./vuid_to_rxcui.csv')
if config.print_status:
print('All API requests for VUIDs complete')
def cui2class(data):
"""
Creates a .csv file containing the results of the RxNorm API
Allows calls to be made only once & can be merged onto data cube
Input: dataframe of unique drug/strength/unit/form along with best-matched RxCUIs
Output: CSV file with the corresponding drug classes
"""
# load data & find unique values
df = data.drop(columns='unique_combo')
search_terms = np.unique(df.values)
# drop missing values
search_terms = search_terms[~np.isnan(search_terms)]
# prep for storing in CSV
cui2class = defaultdict(dict)
if config.print_status:
print('Initiating ' + str(len(search_terms)) +
' API requests for Drug Classes of RxCUIs')
# break dataframe into subsections to prevent API timeout
search_terms = [search_terms[i * config.batch_size:(i + 1) * config.batch_size] \
for i in range((len(search_terms) + config.batch_size - 1) // config.batch_size)]
counter = 0
if config.print_status:
print('Large list of terms. Created ' + str(len(search_terms)) +
' batches.')
for terms in search_terms:
if config.print_status:
counter += 1
print('Starting batch # ' + str(counter))
# make iterable list of URLs to send to api
urls = api_rxnorm.make_url_requests(terms, 'cui2class')
results = api_rxnorm.async_calls(urls)
for result in results:
soup = BeautifulSoup(result, 'html.parser') #'xml')
class_all = soup.find_all('classname')
classes = set([])
for c in class_all:
classes.add(c.get_text())
# add to dictionary
try:
unique_combo = soup.find('rxcui').get_text()
cui2class[unique_combo] = classes
except:
pass
cui2class = pd.DataFrame.from_dict(cui2class, orient='index')
cui2class.reset_index(inplace=True)
# make enough column names for all extra placeholders
cnames = ['rxcui']
for i in range(1, cui2class.shape[1]):
cnames.append('cat' + str(i))
cui2class.columns = cnames
# coerce from dictionary to dataframe to allow merge
#df = pd.DataFrame.from_dict(rxcui_brand_generic, orient='index')
#df.rename(columns={0: 'rxcui1_brand', 1: 'rxcui1_generic'}, inplace=True)
#df['rxcui1'] = df.index.astype(float)
# merge & save
#cui2class = cui2class.merge(df, how='left', on='rxcui1')
cui2class.to_csv(config.out_dir + 'cui2class.csv', index=False)
if config.print_status:
print(
'All API requests for Converting RxCUIs to MESH Classes complete')
def pharmaorder2rxcui():
"""
Creates a .csv file containing the results of the RxNorm API
Allows calls to be made only once & can be merged onto data cube
Input: aggregated source data as a dataframe
Output: CSV file with the unique PharmacyOderableItems that had a corresponding RxNorm CUI
"""
if os.path.exists('./pharmaorder_to_rxcui.csv'):
if config.print_status:
print('PharmacyOrderables already created in .csv file')
return None
# load data & find unique values
df = load_agg_data()
orderables = df.PharmacyOrderableItem.unique()
if config.print_status:
print('Initiating API requests for PharmacyOrderableItems')
# make iterable list of URLs to send to api
urls = api_rxnorm.make_url_requests(orderables, 'string')
results = api_rxnorm.async_calls(urls)
# clean up & store in CSV
pharmaorder_to_rxcui = defaultdict(dict)
for result in results:
soup = BeautifulSoup(result, 'xml')
cui_all = soup.find_all('rxcui')
cui = set([])
for c in cui_all:
cui.add(int(c.get_text()))
# add to dictionary
pharmaorder = soup.find('inputTerm').get_text()
pharmaorder_to_rxcui[pharmaorder] = cui
pharmaorder_to_rxcui = pd.DataFrame.from_dict(pharmaorder_to_rxcui,
orient='index')
pharmaorder_to_rxcui.reset_index(inplace=True)
# make enough column names for all extra placeholders
cnames = ['pharmaorder']
for i in range(1, pharmaorder_to_rxcui.shape[1]):
cnames.append('rxcui' + str(i))
pharmaorder_to_rxcui.columns = cnames
if config.print_status:
print('API requests for PharmacyOrderableItems to RxCUI Complete.')
print('Initiating API requests for generic & brand RxCUIs...')
# search for generic & brand names
rxcui_brand_generic = api_rxnorm.getBrandGeneric(pharmaorder_to_rxcui,
cui_col='rxcui1')
# coerce from dictionary to dataframe to allow merge
df = pd.DataFrame.from_dict(rxcui_brand_generic, orient='index')
df.rename(columns={0: 'rxcui1_brand', 1: 'rxcui1_generic'}, inplace=True)
df['rxcui1'] = df.index.astype(float)
# merge & save
pharmaorder_to_rxcui = pharmaorder_to_rxcui.merge(df,
how='left',
on='rxcui1')
pharmaorder_to_rxcui.to_csv('./pharmaorder_to_rxcui.csv')
if config.print_status:
print('All API requests for PharamcyOrderableItems complete')
def structured_elements2rxcui(data):
"""
Creates a .csv file containing the results of the RxNorm API
Allows calls to be made only once & can be merged onto data cube
Input: aggregated source data as a dataframe
Output: CSV file with the unique combination of ingredient/strength/unit/form that had a corresponding RxNorm CUI
"""
# load data & find unique values
df = data
structured_elements = df[[
'DrugNameWithoutDose', 'StrengthText', 'DrugUnit', 'DosageForm'
]].drop_duplicates()
search_terms = api_rxnorm.make_string(structured_elements)
if config.print_status:
print('Initiating ' + str(len(search_terms)) +
' API requests for Unique Meds')
# prepare for storing in CSV
structured_elements_to_rxcui = defaultdict(dict)
# break dataframe into subsections to prevent API timeout
search_terms = [search_terms[i * config.batch_size:(i + 1) * config.batch_size] \
for i in range((len(search_terms) + config.batch_size - 1) // config.batch_size)]
counter = 0
if config.print_status:
print('Large list of terms. Created ' + str(len(search_terms)) +
' batches.')
for terms in search_terms:
if config.print_status:
counter += 1
print('Starting batch # ' + str(counter))
# make iterable list of URLs to send to api
urls = api_rxnorm.make_url_requests(terms, 'string')
results = api_rxnorm.async_calls(urls)
for result in results:
soup = BeautifulSoup(result, 'html.parser') #'xml')
cui_all = soup.find_all('rxcui')
cui = set([])
for c in cui_all:
cui.add(int(c.get_text()))
# add to dictionary
try:
unique_combo = soup.find('inputterm').get_text()
structured_elements_to_rxcui[unique_combo] = cui
except:
pass
structured_elements_to_rxcui = pd.DataFrame.from_dict(
structured_elements_to_rxcui, orient='index')
structured_elements_to_rxcui.reset_index(inplace=True)
# make enough column names for all extra placeholders
cnames = ['unique_combo']
for i in range(1, structured_elements_to_rxcui.shape[1]):
cnames.append('rxcui' + str(i))
structured_elements_to_rxcui.columns = cnames
if config.print_status:
print('API requests for Structured Elements to RxCUI Complete.')
# merge & save
#structured_elements_to_rxcui = structured_elements_to_rxcui.merge(df, how='left', on='rxcui1')
structured_elements_to_rxcui.to_csv(config.out_dir +
'structured_elements_to_rxcui.csv',
index=False)
if config.print_status:
print('All API requests for Structured Elements complete')
def ndc2rxcui():
"""
Creates a .csv file containing the results of the RxNorm API
Allows calls to be made only once & can be merged onto data cube
Input: aggregated source data as a dataframe
Output: CSV file with the unique NDCs that had a corresponding RxNorm CUI
"""
if os.path.exists('./ndc_to_rxcui.csv'):
if config.print_status:
print('NDCs already created in .csv file')
return None
# load data & find unique values
df = load_agg_data()
ndcs = df.NDC.unique()
if config.print_status:
print('Initiating API requests for NDCs')
# make iterable list of URLs to send to api
urls = api_rxnorm.make_url_requests(ndcs, 'ndc')
results = api_rxnorm.async_calls(urls)
# clean up & store in CSV
ndc_to_rxcui = defaultdict(list)
for result in results:
# extract digits numeric results
try:
# find first instance of any NDC number (assumes one is always present)
pat = re.compile(b'\d+(-\d+)*')
ndc = pat.search(result)[0].decode("utf-8")
# find the corresponding RxNorm CUI
# assumes only 1 RxCUI present & that it falls after the NDC (which is always present)
pat = re.compile(b'\d+')
cui = int(pat.search(result, 47)[0])
except:
continue
# add to dictionary
ndc_to_rxcui[ndc] = cui
ndc_to_rxcui = pd.DataFrame.from_dict(ndc_to_rxcui, orient='index')
ndc_to_rxcui.reset_index(inplace=True)
ndc_to_rxcui.columns = ['ndc', 'rxcui']
if config.print_status:
print('API requests for NDC to RxCUI Complete.')
print('Initiating API requests for generic & brand RxCUIs...')
# search for generic & brand names
rxcui_brand_generic = api_rxnorm.getBrandGeneric(ndc_to_rxcui,
cui_col='rxcui')
# coerce from dictionary to dataframe to allow merge
df = pd.DataFrame.from_dict(rxcui_brand_generic, orient='index')
df.rename(columns={0: 'rxcui_brand', 1: 'rxcui_generic'}, inplace=True)
df['rxcui'] = df.index.astype(float)
# merge & save
ndc_to_rxcui = ndc_to_rxcui.merge(df, how='left', on='rxcui')
ndc_to_rxcui.to_csv('./ndc_to_rxcui.csv')
if config.print_status:
print('All API requests for NDCs complete')