Imputation - Tutorial and Instructions
Introduction
Logic for imputation
The algorithms require users to provide a dataframe of prescriptions (either available directly from EHR databases or extracted with T-Rx modules), with dosage and quantity information extracted for the majority of prescriptions.
The module then splits the prescription dataframe into:
- Reference dataframe: contains prescriptions with extracted/cleaned dosage or quantity information
- Imputation dataframe: contains prescriptions without dosage or quantity information
Most commonly occurred dosage/quantity for every drug present in the reference dataframe is used to impute the information in imputation dataframe, with the same drug name.
By using the original provided dataframe as training dataset for imputation, it also allows the flexibility to capture differences in practices, where the most commonly prescribed dose/strength could differ across regions.
The use of external references for the imputation of dosage and quantity is still under development. Please kindly contact Chris Lo (chris.lowh@kcl.ac.uk) for details.
Figure 1: Overview of the imputation module
Sample dataframe
In this tutorial, we utilized a sample dataframe of uncleaned antidepressant prescriptions in UK Biobank (UKB) primary care records (5000 rows, 4 columns), containing the following columns:
Column name descriptions
| Column | Details | Note |
|---|---|---|
drug_name |
Drug name (from raw UKB primary care records) | Please refer to UKB documentation |
quantity |
Quantity issued (from raw UKB primary care records) | Please refer to UKB documentation |
chem_name |
Drug name of active ingredient | Extracted from
https://github.com/chiarafabbri/MDD_TRD_study |
func_class |
Drug class of active ingredient | Extracted from
https://github.com/chiarafabbri/MDD_TRD_study |
Inspect prescription sample
| drug_name | quantity | chem_name | func_class |
|---|---|---|---|
| Seroxat 20mg tablets (GlaxoSmithKline UK Ltd) | 30 tab - 20 mg | paroxetine | SSRI |
| Dutonin 100mg tablets (Bristol-Myers Squibb Pharmaceuticals Ltd) | 84 tabs | nefazodone | SARI |
| Paroxetine 10mg/5ml oral suspension sugar free | 60 ml(s) - 10 mg/5 ml | paroxetine | SSRI |
| Venlafaxine 37.5mg tablets | 1 - Pack of 56 | venlafaxine | SNRI |
| Clomipramine 25mg capsules | 140 capsules | clomipramine | tricyclic_antidepressant |
| Dosulepin 25mg capsules | 39 capsule(s) | dosulepin | tricyclic_antidepressant |
Imputation of strength / dosage
Initialization
The sample prescriptions were hosted on T-Rx as the data object
antidep_ukb_rx.
Load T-Rx and inspect the data object on R.
# load package
library(TRX)
# inspect sample data object (antidepressant prescriptions)
dim(antidep_ukb_rx)
# [1] 5000 4The sample prescriptions in UK Biobank do not have strength/dosage
information extracted. Run the strength_extract() function
to extract strength/dosage information.
antidep_ukb_extracted = strength_extract(rx_df = antidep_ukb_rx,
liquid_strength_unit = c("mg/5ml"),
solid_strength_unit = c("mg", "mcg", "microgram"),
combined_strength_unit = c("mg", "mcg", "microgram", "miligram"),
info_col = c("drug_name"), combined_strength = 2)Note: For details on
strength_extract(), please refer to the instructions
here.
Resultant dataframe
| drug_name | quantity | chem_name | func_class | strength_unit | strength |
|---|---|---|---|---|---|
| Seroxat 20mg tablets (GlaxoSmithKline UK Ltd) | 30 tab - 20 mg | paroxetine | SSRI | mg | 20 |
| Dutonin 100mg tablets (Bristol-Myers Squibb Pharmaceuticals Ltd) | 84 tabs | nefazodone | SARI | mg | 100 |
| Paroxetine 10mg/5ml oral suspension sugar free | 60 ml(s) - 10 mg/5 ml | paroxetine | SSRI | mg/5ml | 10 |
| Venlafaxine 37.5mg tablets | 1 - Pack of 56 | venlafaxine | SNRI | mg | 37.5 |
| Clomipramine 25mg capsules | 140 capsules | clomipramine | tricyclic_antidepressant | mg | 25 |
| Dosulepin 25mg capsules | 39 capsule(s) | dosulepin | tricyclic_antidepressant | mg | 25 |
Log file (as messages)
Inspecting problematic prescriptions
From the log files, 31 rows of prescriptions did not have strength/dosage information available.
There are also prescriptions with multiple strength, the dosage of
which were separated by ,. However, these strengths need to
be matched to specific drugs within the multi-strength products and
expanded into separate rows.
Missing strength/dosage information
| drug_name | quantity | chem_name | func_class | strength_unit | strength | |
|---|---|---|---|---|---|---|
| 206 | NEFAZODONE starter pack | 56.000 | nefazodone | SARI | NA | NA |
| 313 | Triptafen tablets (AMCo) | 400 tablet(s) | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 363 | Motipress tablets (Sanofi-Synthelabo Ltd) | 30 tablets | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 556 | Triptafen tablets (AMCo) | tablet(s) | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 591 | NEFAZODONE STARTER PACK [TABS (STARTER PACK)] | 1.000 | nefazodone | SARI | NA | NA |
| 661 | PARSTELIN tablets [GLAXSK CON] | 84 | tranylcypromine | MAOI | NA | NA |
| 1172 | TRIPTAFEN-M TAB | 120.000 | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 1180 | Triptafen tablets (AMCo) | 30 tabs | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 1254 | TRIPTAFEN | 56.000 | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 1329 | TRIPTAFEN TAB | 60.000 | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 1363 | MOTIVAL tabs | 60.000 | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 1592 | Triptafen tablets (AMCo) | 120.000 | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 1641 | Motipress tablets (Sanofi-Synthelabo Ltd) | tablet(s) | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 1994 | Trazodone Controlled Rele | trazodone | SARI | NA | NA | |
| 2374 | Motival TABS | 112.000 | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 2498 | Triptafen tablets (AMCo) | 30 tablet(s) | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 2793 | Triptafen tablets (AMCo) | 2*100 tablet(s) | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 2882 | nefazodone starter pack | 56 tablets | nefazodone | SARI | NA | NA |
| 3053 | Triptafen tablets (AMCo) | 14 tablets | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 3187 | nefazodone starter pack | tablet(s) - 14 x 50 mg, 14 x 100 mg, 28 x 200 mg | nefazodone | SARI | NA | NA |
| 3404 | PARSTELIN tablets [GLAXSK CON] | x 60 | tranylcypromine | MAOI | NA | NA |
| 3676 | MOTIVAL TABLETS | 50.000 | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 3689 | NEFAZODONE starter pack | 1.000 | nefazodone | SARI | NA | NA |
| 3848 | Triptafen tablets (AMCo) | 240 tablet(s) | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 4125 | Triptafen tablets (AMCo) | 30.000 | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | NA | NA |
| 4387 | CITALOPRAM | 28.000 | citalopram | SSRI | NA | NA |
| 4392 | nefazodone starter pack | 56 tabs (starter pack) | nefazodone | SARI | NA | NA |
| 4516 | SERTRALINE | 56.000 | sertraline | SSRI | NA | NA |
| 4535 | nefazodone starter pack | 1 - tablets (14x50mg,14x100mg,28x200mg) | nefazodone | SARI | NA | NA |
| 4668 | Dutonin tablets treatment initiation pack (Bristol-Myers Squibb Pharmaceuticals Ltd) | 56 | nefazodone | SARI | NA | NA |
| 4782 | Parstelin Tablet (GlaxoSmithKline Consumer Healthcare) | 112.000 | tranylcypromine_trifluoperazine | MAOI_typical_antipsychotic | NA | NA |
Multiple strength products
| drug_name | quantity | chem_name | func_class | strength_unit | strength | |
|---|---|---|---|---|---|---|
| 72 | TRIPTAFEN M tablets 2mg + 10mg [AMCO] | 30 | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 327 | Motival 10mg/500microgram tablets (Sanofi) | 3 op | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,microgram | 10,500 |
| 403 | perphenazine with amitriptyline tablets 2mg + 25mg | 28 - tablet(s) | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,25 |
| 429 | TRIPTAFEN M tablets 2mg + 10mg [AMCO] | 120 tablets | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 498 | perphenazine with amitriptyline tablets 2mg + 10mg | 56 - tablet(s) | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 571 | Motival 10mg/500microgram tablets (Sanofi) | 30.000 | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,microgram | 10,500 |
| 663 | TRIPTAFEN M tablets 2mg + 10mg [AMCO] | 168 tablet | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 883 | Motival 10mg/500microgram tablets (Sanofi) | 60 tablets | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,microgram | 10,500 |
| 995 | TRIPTAFEN M tablets 2mg + 10mg [AMCO] | 30 tab | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 1102 | Motival 10mg/500microgram tablets (Sanofi) | 90 | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,microgram | 10,500 |
| 1341 | Motival 10mg/500microgram tablets (Sanofi) | 168 | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,microgram | 10,500 |
| 1377 | TRIPTAFEN M tablets 2mg + 10mg [AMCO] | 42 | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 1431 | Motival 10mg/500microgram tablets (Sanofi) | 28 tablets | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,microgram | 10,500 |
| 1539 | Motival 10mg/500microgram tablets (Sanofi) | 84 tabs | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,microgram | 10,500 |
| 1581 | Motival 10mg/500microgram tablets (Sanofi) | 120 | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,microgram | 10,500 |
| 1634 | perphenazine with amitriptyline tablets 2mg + 25mg | 56 tablets | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,25 |
| 1705 | Triptafen m 2mg+10mg Tablet (Goldshield Pharmaceuticals Ltd) | 84.000 | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 1810 | Motival 10mg/500microgram tablets (Sanofi) | 90 tablets | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,microgram | 10,500 |
| 1899 | TRIPTAFEN M tabs 2mg + 10mg | 84.000 | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 2045 | TRIPTAFEN M tablets 2mg + 10mg [AMCO] | 30 tablets | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 2079 | perphenazine with amitriptyline tablets 2mg + 25mg | 100 - tablet(s) | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,25 |
| 2281 | Motival 10mg/500microgram tablets (Sanofi) | 60 tabs | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,microgram | 10,500 |
| 2695 | Motival 10mg/500microgram tablets (Sanofi) | 56 tab | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,microgram | 10,500 |
| 2817 | Motival 10mg/500microgram tablets (Sanofi) | 60 - tablet(s) | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,microgram | 10,500 |
| 3301 | TRIPTAFEN M tablets 2mg + 10mg [AMCO] | 112 tablet(s) | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 3862 | TRIPTAFEN M tablets 2mg + 10mg [AMCO] | 40 | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 3906 | Motival 10mg/500microgram tablets (Sanofi) | 30 tablet(s) | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,microgram | 10,500 |
| 3942 | TRIPTAFEN M tabs 2mg + 10mg | 168.000 | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 4046 | TRIPTAFEN M tablets 2mg + 10mg [AMCO] | 28 tablets | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 4300 | PERPHENAZINE + AMITRIPTYLINE tabs 2mg + 10mg | 84.000 | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 4391 | TRIPTAFEN M tablets 2mg + 10mg [AMCO] | 56 | amitriptyline_perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,mg | 2,10 |
| 4928 | Motival 10mg/500microgram tablets (Sanofi) | 100 tablets | nortriptyline_fluphenazine | tricyclic_antidepressant_typical_antipsychotic | mg,microgram | 10,500 |
strength_impute()
Logic
strength_impute() primarily performs two actions:
Creating an imputation reference, then match to prescriptions without strength/dosage information;
Expanding multiple strength products (assuming they were contained in single row with separator strings) to prescriptions of multiple rows, each representing the correct dosage of individual ingredients within the product.
Note: Expansion of multiple strength products builds on the presumption that prescriptions of the multi-strength product is summarised in one single row, with information on the names, strengths and strength units of each constituent drug.
Figure 2: The strength_impute() function
Instructions
Using the above antidep_ukb_extracted dataframe, run the
strength_impute() function.
impute_df = strength_impute(rx_df = antidep_ukb_extracted,
drug_separator = "_", ref_drug_col = "chem_name",
strength_separator = ",", strength_colname = "strength",
strength_unit_colname = "strength_unit", strength_unit_separator = ",")Note: If you wish to only perform imputation without
the need to expand multiple strength products, please still provide the
drug_separator, strength_separator and
strength_unit_separator arguments as default. It would not
affect imputation algorithms unless the drug names contains strings of
the separator specified.
Description on arguments
| Argument | Details |
|---|---|
rx_df |
Prescription dataframe |
drug_separator |
The separator used to separate multiple drug names in a single cell |
ref_drug_col |
The column name of the drug name in the dataframe |
strength_separator |
The separator used to separate multiple strength values in a single cell |
strength_colname |
The column name for strength information |
strength_unit_separator |
The separator used to separate multiple strength unit values in a single cell |
strength_unit_colname |
The column name for strength units |
Log file (as messages)
Output
The log file shows:
- 52 prescriptions with multi-ingredient products (scanned based on the separator arguments)
- 31 prescriptions requires strength imputation
(
<NA>instrength_colname)
Here we list out prescription samples of missing strengths and multiple ingredients for inspection.
Details for imputed prescriptions
Prescription with missing strength (Sertraline as example)
| drug_name | quantity | chem_name | func_class | strength_unit | strength | |
|---|---|---|---|---|---|---|
| 4566 | SERTRALINE | 56.000 | sertraline | SSRI | mg | 50 |
Prescription with multiple strengths
| drug_name | quantity | chem_name | func_class | strength_unit | strength |
|---|---|---|---|---|---|
| perphenazine with amitriptyline tablets 2mg + 25mg | 28 - tablet(s) | perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg | 2 |
| perphenazine with amitriptyline tablets 2mg + 25mg | 28 - tablet(s) | amitriptyline | tricyclic_antidepressant_typical_antipsychotic | mg | 25 |
You would see the same row of prescription is now expanded to two rows, which contains perphenazine and amitriptyline with the strength/dosage mapped to the correct drugs.
Failure for imputation
Out of the 31 prescriptions, 30 can be imputed. The remaining
prescriptions which the dosage cannot be imputed, is coded as
failure string instead.
| drug_name | quantity | chem_name | func_class | strength_unit | strength | |
|---|---|---|---|---|---|---|
| 4833 | Parstelin Tablet (GlaxoSmithKline Consumer Healthcare) | 112.000 | trifluoperazine | MAOI_typical_antipsychotic | failure | failure |
Imputation of quantity
COMING SOON
This section is currently under development. Please stay tuned and check later!
Additional functions for UKB primary care records
multi_num_infer()
Logic
The multi_num_infer() function is a function to infer
quantity information from UKB-formatted records. Please use with caution
when used in other settings as this function is currently not validated
for prescription records outside of the UKB.
In UKB-formatted records, most details of the prescriptions were
described in the drug_name and quantity
columns.
The two columns usually contain information on:
- Dosage / Strength
- Quantity
- Multiplier
Therefore, if the dosage, multiplier information were extracted as integers already, we could infer the remaining integer to be quantity.
To achieve this, the multi_num_infer() function looks
for all integers that were present in info_col, and compare
the list of integers to the existing extracted integers in strength and
multiplier information.
If there is only one outstanding integer, we can infer that as the
quantity. Otherwise, the quantity would remain as
<NA> (as ambiguity in extraction).
Figure 3: The multi_num_infer() function
Initialization
The multi_num_infer() function works when
dosage/strength, quantity and multiplier information were extracted
(fully or partly) using upstream functions in T-Rx, including
strength_extract()(with option to runstrength_impute())quantity_extract()multiplier_extract()
Using antidepressant prescriptions in UK Biobank as an example
(antidep_ukb_rx), run the above functions to clean the
prescription records.
Functions
# Run strength extraction function with UK Biobank prescription sample
antidep_ukb_infer = strength_extract(rx_df = antidep_ukb_rx,
liquid_strength_unit = c("mg/ml","mg/5ml"),
solid_strength_unit = c("mg", "mcg", "microgram"),
combined_strength_unit = c("mg", "mcg", "microgram", "miligram"),
info_col = c("drug_name"), combined_strength = 2,
no_strength_unit_remove = FALSE, no_strength_unit = "mg")
# Run strength imputation function
antidep_ukb_infer = strength_impute(rx_df = antidep_ukb_infer,
drug_separator = "_", ref_drug_col = "chem_name",
strength_separator = ",", strength_colname = "strength",
strength_unit_colname = "strength_unit", strength_unit_separator = ",")
# remove prescription with failure in imputation
antidep_ukb_infer = antidep_ukb_infer[!grepl("failure", antidep_ukb_infer$strength),]
# Run quantity extraction function
antidep_ukb_infer = quantity_extract(rx_df = antidep_ukb_infer,
liquid_forms = c("susp", "suspension", "liq", "liquid", "syrup", "soln", "solution", "drop", "elixir"),
solid_forms = c("tab", "cap"),
form_info_col = c("quantity", "drug_name"),
dose_form_colname = "dosage_form",
form_quant_colname = "form_quantity",
liq_form_suffix = c("ml", "millilitre"),
quantity_info_col = c("quantity"), no_form_remove = TRUE)
# Run multiplier extraction function
antidep_ukb_infer = multiplier_extract(rx_df = antidep_ukb_infer,
multipliers = c("pack", "pack of" , "\\*"),
alt_multipliers = c("x"),
info_col = c("quantity"),
multiplier_colname = "multiplier", qc_remove = TRUE)Note: We inserted quality control parameters for
quantity_extract() and multiplier_extract() to
make up a cleaned presccription dataframe. Please refer to the
instructions here for
details.
Inspecting problematic prescriptions, with missing quantity (first 6 rows)
| drug_name | quantity | chem_name | func_class | strength_unit | strength | dosage_form | form_quantity | multiplier | |
|---|---|---|---|---|---|---|---|---|---|
| 4 | Venlafaxine 37.5mg tablets | 1 - Pack of 56 | venlafaxine | SNRI | mg | 37.5 | tab | NA | 1 |
| 9 | Duloxetine 30mg gastro-resistant capsules | 60.000 | duloxetine | SNRI | mg | 30 | cap | NA | 1 |
| 10 | AMITRIPTYLINE 25mg tablets | 60.000 | amitriptyline | tricyclic_antidepressant | mg | 25 | tab | NA | 1 |
| 11 | Mirtazapine Orodispersible TABS 30MG | 90.000 | mirtazapine | other | mg | 30 | tab | NA | 1 |
| 13 | Duloxetine Gastro Resistant CAPS 30MG | 14.000 | duloxetine | SNRI | mg | 30 | cap | NA | 1 |
| 19 | DOTHIEPIN TABS 75MG | 21.000 | dosulepin | tricyclic_antidepressant | mg | 75 | tab | NA | 1 |
Instructions
Using the above half-cleaned dataframe, run the
multi_num_infer() function:
antidep_ukb_infer_post = multi_num_infer(rx_df = antidep_ukb_infer,
info_col = "quantity", strength_colname = "strength",
strength_unit_colname = "strength_unit",
dose_form_colname = "dosage_form",
form_quant_colname = "form_quantity")Description on arguments
| Argument | Details |
|---|---|
rx_df |
Prescription dataframe |
strength_colname |
The column name in ‘rx_df’ containing strength information |
strength_unit_colname |
The column name in ‘rx_df’ containing strength unit information |
dose_form_colname |
The column name in ‘rx_df’ containing dosage form information |
form_quant_colname |
The column name in ‘rx_df’ containing form quantity information |
Output
The multi_num_infer() function provides inferrence on
quantity information on many prescriptions where quantity information
cannot be extracted from quantity_extract().
Run the below lines to check pre-/post-inferrence changes.
# Pre-inferrence
sum(is.na(antidep_ukb_infer$form_quantity)) # 2380
# Post-inferrence
sum(is.na(antidep_ukb_infer_post$form_quantity)) #70Out of the 2380 prescriptions without quantity information extracted, only 70 of them did not have the quantity information inferred.
Checking problematic prescriptions
Run the below lines to check problematic prescriptions where inferrence cannot be made.
Code
antidep_ukb_infer_na = antidep_ukb_infer_post[is.na(antidep_ukb_infer_post$form_quantity),]Prescriptions with missing information on quantity after inferrence
| drug_name | quantity | chem_name | func_class | strength_unit | strength | dosage_form | form_quantity | multiplier | |
|---|---|---|---|---|---|---|---|---|---|
| 58 | Trimipramine 25mg tablets | trimipramine | tricyclic_antidepressant | mg | 25 | tab | NA | 1 | |
| 79 | Amitriptyline 10mg tablets | tablet(s) Tablets | amitriptyline | tricyclic_antidepressant | mg | 10 | tab | NA | 1 |
| 122 | DOTHIEPIN TABS 75 MG | dosulepin | tricyclic_antidepressant | mg | 75 | tab | NA | 1 | |
| 142 | Prozac 20mg capsules (Eli Lilly and Company Ltd) | 2 - Pack of 28 (2 X 14) | fluoxetine | SSRI | mg | 20 | cap | NA | 2 |
| 159 | Fluanxol 1mg tablets (Lundbeck Ltd) | tablet(s) | flupentixol | phenothiazine_antipsychotic | mg | 1 | tab | NA | 1 |
| 193 | Citalopram 20mg tablets | tablet(s) Tablets | citalopram | SSRI | mg | 20 | tab | NA | 1 |
| 213 | Seroxat 20mg tablets (GlaxoSmithKline UK Ltd) | 2 - Pack of 30 (2 X 15) | paroxetine | SSRI | mg | 20 | tab | NA | 2 |
| 311 | Escitalopram 10mg tablets | See Dosage For Quantity | escitalopram | SSRI | mg | 10 | tab | NA | 1 |
| 531 | Sertraline 100mg tablets | tablet(s) | sertraline | SSRI | mg | 100 | tab | NA | 1 |
| 562 | Triptafen tablets (AMCo) | tablet(s) | amitriptyline | tricyclic_antidepressant_typical_antipsychotic | mg | 10 | tab | NA | 1 |
| 563 | Triptafen tablets (AMCo) | tablet(s) | perphenazine | tricyclic_antidepressant_typical_antipsychotic | mg | 2 | tab | NA | 1 |
| 719 | Mirtazapine 30mg tablets | -1 - 30 mg | mirtazapine | other | mg | 30 | tab | NA | 1 |
| 802 | Mirtazapine 30mg tablets | See Dosage For Quantity | mirtazapine | other | mg | 30 | tab | NA | 1 |
| 805 | Dosulepin 25mg capsules (Almus Pharmaceuticals Ltd) | See Dosage For Quantity | dosulepin | tricyclic_antidepressant | mg | 25 | cap | NA | 1 |
| 1082 | Fluoxetine 20mg capsules | 3 - Pack of 30 (2 X 15) | fluoxetine | SSRI | mg | 20 | cap | NA | 3 |
| 1245 | Citalopram tablets 20mg | citalopram | SSRI | mg | 20 | tab | NA | 1 | |
| 1353 | Sertraline 100mg tablets | 1 op - 100 mg | sertraline | SSRI | mg | 100 | tab | NA | 1 |
| 1619 | Prozac 20mg capsules (Eli Lilly and Company Ltd) | capsule(s) Capsules | fluoxetine | SSRI | mg | 20 | cap | NA | 1 |
| 1660 | Motipress tablets (Sanofi-Synthelabo Ltd) | tablet(s) | nortriptyline | tricyclic_antidepressant_typical_antipsychotic | mg | 10 | tab | NA | 1 |
| 1661 | Motipress tablets (Sanofi-Synthelabo Ltd) | tablet(s) | fluphenazine | tricyclic_antidepressant_typical_antipsychotic | microgram | 500 | tab | NA | 1 |
| 1688 | Amitriptyline 10mg tablets | amitriptyline | tricyclic_antidepressant | mg | 10 | tab | NA | 1 | |
| 1741 | Dosulepin 75mg tablets | 1 months supply - 75 mg | dosulepin | tricyclic_antidepressant | mg | 75 | tab | NA | 1 |
| 1926 | Mianserin 10mg tablets | mianserin | tetracyclic_antidepressant | mg | 10 | tab | NA | 1 | |
| 1996 | Fluoxetine 20mg capsules | 1 month - 20 mg | fluoxetine | SSRI | mg | 20 | cap | NA | 1 |
| 2026 | Molipaxin 150mg tablets (Zentiva) | trazodone | SARI | mg | 150 | tab | NA | 1 | |
| 2092 | Amitriptyline 25mg tablets | one month | amitriptyline | tricyclic_antidepressant | mg | 25 | tab | NA | 1 |
| 2166 | Citalopram 10mg tablets | 1 28x3 | citalopram | SSRI | mg | 10 | tab | NA | 28 |
| 2203 | Lustral 100mg tablets (Pfizer Ltd) | See Dosage For Quantity | sertraline | SSRI | mg | 100 | tab | NA | 1 |
| 2254 | Citalopram 20mg tablets | citalopram | SSRI | mg | 20 | tab | NA | 1 | |
| 2358 | Mianserin Hydrochloride Tablets 30 mg | mianserin | tetracyclic_antidepressant | mg | 30 | tab | NA | 1 | |
| 2458 | Cipramil 20mg tablets (Lundbeck Ltd) | 4/52 | citalopram | SSRI | mg | 20 | tab | NA | 1 |
| 2470 | Prothiaden 25mg capsules (Teofarma) | 4/52 | dosulepin | tricyclic_antidepressant | mg | 25 | cap | NA | 1 |
| 2473 | CLOMIPRAMINE capsules 50mg [SANDOZ] | clomipramine | tricyclic_antidepressant | mg | 50 | cap | NA | 1 | |
| 2533 | Venlafaxine 37.5mg tablets | See Dosage For Quantity | venlafaxine | SNRI | mg | 37.5 | tab | NA | 1 |
| 2612 | Lofepramine 70mg tablets | See Dosage For Quantity | lofepramine | tricyclic_antidepressant | mg | 70 | tab | NA | 1 |
| 2643 | Seroxat 20mg tablets (GlaxoSmithKline UK Ltd) | tablet(s) Tablets | paroxetine | SSRI | mg | 20 | tab | NA | 1 |
| 2670 | Surmontil 50mg capsules (Sanofi) | 1 o.p. - 50 mg | trimipramine | tricyclic_antidepressant | mg | 50 | cap | NA | 1 |
| 2693 | MARPLAN TAB 10mg | isocarboxazid | MAOI | mg | 10 | tab | NA | 1 | |
| 2894 | Amitriptyline 25mg tablets | EIGHTY FOUR | amitriptyline | tricyclic_antidepressant | mg | 25 | tab | NA | 1 |
| 3071 | Maprotiline 25mg tablets | See Dosage For Quantity | maprotiline | tetracyclic_antidepressant | mg | 25 | tab | NA | 1 |
| 3084 | Oxactin 20mg capsules (Discovery Pharmaceuticals) | 1 30 x 2 | fluoxetine | SSRI | mg | 20 | cap | NA | 30 |
| 3108 | Lustral 50mg tablets (Pfizer Ltd) | tablets | sertraline | SSRI | mg | 50 | tab | NA | 1 |
| 3155 | Amitriptyline 25mg tablets | 1 o.p - 25 mg | amitriptyline | tricyclic_antidepressant | mg | 25 | tab | NA | 1 |
| 3167 | Fluoxetine 20mg capsules | 3 Pack of 30 (2 X 15) | fluoxetine | SSRI | mg | 20 | cap | NA | 3 |
| 3223 | Maprotiline 50mg tablets | See Dosage For Quantity | maprotiline | tetracyclic_antidepressant | mg | 50 | tab | NA | 1 |
| 3251 | Dosulepin 75mg tablets | tablet(s) Tablets | dosulepin | tricyclic_antidepressant | mg | 75 | tab | NA | 1 |
| 3257 | Lofepramine 70mg tablets | 2 /52 | lofepramine | tricyclic_antidepressant | mg | 70 | tab | NA | 1 |
| 3266 | Citalopram 20mg tablets | -1 - 20 mg | citalopram | SSRI | mg | 20 | tab | NA | 1 |
| 3319 | Reboxetine 4mg tablets | 1 pack - 4 mg | reboxetine | NRI | mg | 4 | tab | NA | 1 |
| 3378 | Prothiaden 25mg capsules (Teofarma) | dosulepin | tricyclic_antidepressant | mg | 25 | cap | NA | 1 | |
| 3479 | Manerix 150mg tablets (Meda Pharmaceuticals Ltd) | tablets | moclobemide | MAOI | mg | 150 | tab | NA | 1 |
| 3648 | Lentizol 25mg modified-release capsules (Pfizer Ltd) | amitriptyline | tricyclic_antidepressant | mg | 25 | cap | NA | 1 | |
| 3651 | Dosulepin 75mg tablets | tablet(s) | dosulepin | tricyclic_antidepressant | mg | 75 | tab | NA | 1 |
| 3654 | Paroxetine 20mg tablets | paroxetine | SSRI | mg | 20 | tab | NA | 1 | |
| 3696 | Paroxetine 30mg tablets | paroxetine | SSRI | mg | 30 | tab | NA | 1 | |
| 3738 | Nortriptyline 10mg tablets | nortriptyline | tricyclic_antidepressant | mg | 10 | tab | NA | 1 | |
| 3789 | Fluvoxamine 50mg tablets | fluvoxamine | SSRI | mg | 50 | tab | NA | 1 | |
| 3840 | Mianserin Hydrochloride Tablets 10 mg | mianserin | tetracyclic_antidepressant | mg | 10 | tab | NA | 1 | |
| 3889 | Seroxat 30mg tablets (GlaxoSmithKline UK Ltd) | See Dosage For Quantity | paroxetine | SSRI | mg | 30 | tab | NA | 1 |
| 4003 | Venlafaxine 150mg modified-release capsules | venlafaxine | SNRI | mg | 150 | cap | NA | 1 | |
| 4348 | Seroxat 20mg tablets (GlaxoSmithKline UK Ltd) | 1 months supply - 20 mg | paroxetine | SSRI | mg | 20 | tab | NA | 1 |
| 4505 | Prozac 20mg capsules (Eli Lilly and Company Ltd) | fluoxetine | SSRI | mg | 20 | cap | NA | 1 | |
| 4600 | Citalopram 40mg tablets | 4/52 | citalopram | SSRI | mg | 40 | tab | NA | 1 |
| 4617 | Lustral 50mg tablets (Pfizer Ltd) | tablet(s) | sertraline | SSRI | mg | 50 | tab | NA | 1 |
| 4646 | Lofepramine 70mg tablets | 2 Pack of 56 (4 X 14) | lofepramine | tricyclic_antidepressant | mg | 70 | tab | NA | 2 |
| 4879 | Paroxetine 20mg tablets | 2 - Pack of 30 (2 X 15) | paroxetine | SSRI | mg | 20 | tab | NA | 2 |
| 4895 | SERTRALINE tabs 50mg | sertraline | SSRI | mg | 50 | tab | NA | 1 | |
| 4922 | Lustral 50mg tablets (Pfizer Ltd) | 1 Pack of 28 (2 X 14) | sertraline | SSRI | mg | 50 | tab | NA | 1 |
| 4992 | PROTHIADEN tablets 75mg | dosulepin | tricyclic_antidepressant | mg | 75 | tab | NA | 1 | |
| 4995 | Mirtazapine 15mg tablets | mirtazapine | other | mg | 15 | tab | NA | 1 |
The remainder prescriptions without quantity information primarily consist of:
- Genuine missingness in quantity
- Quantity summarised as duration (e.g.:
4/52) - Conflicting numbers (e.g.:
2 Pack of 56 (4 X 14))
It is recommended to handle these prescriptions manually, or use some
of the downstream functions (such as duration_handling(),
quantity_impute()).
Optional Arguments
If users wish to remove the prescriptions without quantity
information inferred, change ambig_filter to
TRUE.
antidep_ukb_infer_post = multi_num_infer(rx_df = antidep_ukb_infer,
info_col = "quantity", strength_colname = "strength",
strength_unit_colname = "strength_unit",
dose_form_colname = "dosage_form",
form_quant_colname = "form_quantity",
ambig_filter = TRUE)If the ambig_filter argument is set to
TRUE, prescriptions without quantity information inferred
would be removed.
duration_handling()
Logic
The duration_handling() function is a function to infer
quantity information from duration-based strings.
Please use with caution when used in other settings as this function is currently not validated for prescription records outside of the UKB.
Prescriptions were sometimes written as duration-based strings, corresponding as the following:
| Duration | Strings |
|---|---|
months |
/12, month, months |
weeks |
/52, week, weeks |
days |
/365, day, days |
The duration_handling() function handles these strings
and convert them into quantity.
Figure 4: The duration_handling() function
Initialization
It is recommended to run duration_handling() function
when the users are certain that there are no other combinations of
strings outside of the table of duration-based strings listed above.
We used antidep_ukb_infer_post and
antidep_ukb_infer_na dataframe as starting points, which
has got dosage and quantity information inferred for most
prescriptions.
Please refer to multi_num_infer() sections for
details.
antidep_ukb_infer_na = antidep_ukb_infer_post[is.na(antidep_ukb_infer_post$form_quantity),]Instructions
For simplicity, run the duration_handling() function on
antidep_ukb_infer_na dataframe, where quantity information
cannot be imputed from multi_num_infer().
antidep_ukb_infer_na = duration_handling(rx_df = antidep_ukb_infer_na,
info_col = c("quantity"), form_quant_colname = c("form_quantity"))The function looks for duration-based strings in the
columns listed in the info_col argument, but only for the
prescriptions without pre-existing quantity information (specified using
form_quantity argument).
Description on arguments
| Argument | Details |
|---|---|
rx_df |
Prescription dataframe |
month_string |
A character string or vector of strings representing months |
week_string |
A character string or vector of strings representing weeks |
day_string |
A character string or vector of strings representing days |
info_col |
The column names in ‘rx_df’ to search for duration-based strings |
form_quant_colname |
The column name in ‘rx_df’ where form quantities are stored |
Note: We have configured default strings for
duration-based strings from the above table. If users wish to change the
strings for search, change the month_string,
week_string or day_string arguments.
Output (only showing prescriptions with duration-based strings)
| drug_name | quantity | chem_name | func_class | strength_unit | strength | dosage_form | form_quantity | multiplier | |
|---|---|---|---|---|---|---|---|---|---|
| 22 | Dosulepin 75mg tablets | 1 months supply - 75 mg | dosulepin | tricyclic_antidepressant | mg | 75 | tab | 30 | 1 |
| 24 | Fluoxetine 20mg capsules | 1 month - 20 mg | fluoxetine | SSRI | mg | 20 | cap | 30 | 1 |
| 31 | Cipramil 20mg tablets (Lundbeck Ltd) | 4/52 | citalopram | SSRI | mg | 20 | tab | 28 | 1 |
| 32 | Prothiaden 25mg capsules (Teofarma) | 4/52 | dosulepin | tricyclic_antidepressant | mg | 25 | cap | 28 | 1 |
| 47 | Lofepramine 70mg tablets | 2 /52 | lofepramine | tricyclic_antidepressant | mg | 70 | tab | 14 | 1 |
| 61 | Seroxat 20mg tablets (GlaxoSmithKline UK Ltd) | 1 months supply - 20 mg | paroxetine | SSRI | mg | 20 | tab | 30 | 1 |
| 63 | Citalopram 40mg tablets | 4/52 | citalopram | SSRI | mg | 40 | tab | 28 | 1 |
Troubleshooting
Please post questions as an issue on the T-Rx GitHub repo here.
The T-Rx package is currently under beta testing. Most functions should have adequate documentation on possible errors.
Please kindly reach out to Chris Lo (chris.lowh@kcl.ac.uk) for feedback on documentation.
