The PROscorer package is an extensible repository of functions to score specific patient-reported outcome (PRO), quality of life (QoL), and other psychometric measures and questionnaire-based instruments commonly used in research.
(Note: For simplicity, from here forward I will collectively and somewhat imprecisely refer to these types of instruments as “PRO measures”, “PRO-like instruments”, or just “PROs”.)
Recent efforts by International Society for Quality of Life Research (ISOQOL) taskforces have sought to standardize the reporting, analysis, and protocol descriptions of PRO measures. However, no best practice guidance nor standardized software exists for scoring PROs. The PROscorer project was designed to establish best practices for scoring PROs by providing a system to standardize the scoring and documentation of commonly-used PROs. Importantly, the PROscorer R package also facilitates the integration of PRO scoring into scientifically reproducible workflows.
Each function in the PROscorer package scores a different
PRO measure. Functions are named using the initials of the PRO measure.
For example, the fsfi function scores the Female Sexual
Function Index (FSFI).
PROscorer also comes with a vignette containing detailed descriptions of each of the instruments scored by PROscorer (see main PROscorer page on CRAN). The purpose of including these instrument descriptions, complete with references, is to help improve the descriptions of PRO measures in protocols, grants, and published results. In most cases, the descriptions can be used in research documents with little or no editing.
To minimize the possibility of scoring errors and other bugs, each PROscorer function is composed of simpler, well-tested “helper” functions from the PROscorerTools package. This reliance on a small set of simple functions that have been thoroughly tested ensures that the underlying code base of PROscorer functions is bug-free, and that the scoring functions produce reliable, consistent, and accurate results.
PROscorer, together with the PROscorerTools package, is a system to facilitate the incorporation of PRO measures into research studies and clinical settings in a scientifically rigorous and reproducible manner. The overarching goals of the PROscorer and PROscorerTools packages are to draw attention to best-practices for PRO scoring and reporting, and to help eliminate inaccurate and inconsistent scoring by standardizing the scoring procedures for commonly used PRO measures.
The scientific rigor and reproducibility of research involving PRO, QoL, and similar measures is lagging behind other research areas. Three major reasons for these shortcomings are (1) measurement error introduced by faulty scoring procedures, (2) inconsistent application of scoring instructions across different studies using the same PRO measures, and (3) inadequate, incomplete, and/or inaccurate descriptions of PRO-like measures in research protocols and in published results of studies that incorporate such measures.
Scoring procedures represent a major source of error in research studies that rely upon PRO and similar measures. These errors typically go unnoticed, hidden, and/or ignored, eroding the scientific integrity of the research and hindering progress in the numerous scientific fields that conduct studies that use these measures.
Similarly, inconsistent application of PRO scoring procedures and variation in scoring across studies makes study results less likely to replicate and slows the accumulation of reliable scientific data from the PRO measure.
Inadequate, incomplete, and/or inaccurate descriptions of PRO-like measures in research documents can cause confusion and introduce errors, oversights, and other mistakes at multiple stages in the research process.
The PROscorer package provides a standardized framework for addressing these problems with research involving PRO-like measures. The lofty goal of the PROscorer package is to eliminate these serious deficiencies in PRO-based research by serving as the gold-standard open-source repository of scoring syntax and instrument descriptions for PRO-like measures commonly used in research and clinical settings.
The features of the PROscorer package and supporting infrastructure were carefully planned with this ambitious goal in mind.
PROscorer serves as the repository of scoring functions for specific, commonly-used PRO measures (e.g., the EORTC QLQ-C30).
Each function scores a specific PRO instrument (e.g., the
qlq_c30 function scores the EORTC QLQ-C30 quality of life
questionnaire).
Protocol- and manuscript-ready descriptions of each PRO instrument are included in the supplementary documentation (i.e., vignettes).
Each function has thorough, helpful documentation.
Consistent naming conventions are used for functions and function arguments (e.g., functions are named using the initials of the instrument).
Scoring functions require minimal user input.
Functions are composed modularly from a small number of reusable, well-tested “helper” functions. A separate package, PROscorerTools, contains these helper functions.
Advanced users can use PROscorerTools to write functions to score their favorite PROs, and then submit them on GitHub for inclusion in future PROscorer updates.
Makes it easy to incorporate PRO scoring into reproducible scientific workflows.
Install the stable version of PROscorer from CRAN:
Load PROscorer into your R workspace with the following:
As an example, we will use the makeFakeData function
from the PROscorerTools package to make fake item responses to
the EORTC QLQ-C30 quality of life questionnaire. The created data set
(named “dat”) has an “id” variable, plus responses to 30 items (named
“q1”, “q2”, etc.) from 20 imaginary respondents. There are also missing
responses (“NA”) scattered throughout.
Below we will use the qlq_c30 function to score the fake
responses in “dat”. We will save the scores from the EORTC QLQ-C30
questionnaire in a data frame named “c30scores”.
The first argument to qlq_c30 took our data frame,
“dat”. With the second argument, we needed to tell the
qlq_c30 function how to find our items in “dat”. Since our
items are all named with the prefix “q” plus the item number, we gave
this quoted prefix to the second argument. These arguments actually have
names, but in most cases you don’t have to explicitly use the names.
Below gives the same results, but explicitly uses the argument
names.
Specifically, the first argument is named df (for
data frame) and the second is named
iprefix (for item prefix).
If you want to merge your scores back into your main data frame with
the item responses, there are several different ways to do so. For
example, assuming you have not changed the order of dat or
dat_scored, you can do the following:
For more information on the qlq_c30 function, you can
access its “help” page by typing ?qlq_c30 into R.
The PROscorer family of R packages includes PROscorer, PROscorerTools, and FACTscorer. With respect to developing PROscorer, my priorities are:
Expand PROscorer with more scoring functions for specific PROs. Some of the EORTC instruments are high on my list.
Further refine some behind-the-scenes standards for how the functions should be programmed, and write guides for users wishing to program and contribute their own PRO scoring functions to PROscorer.
Finalize the collaborative infrastructure (e.g., on GitHub) by which users can use PROscorerTools to write scoring functions for their favorite PROs and submit them for inclusion in PROscorer. A major component of this is to create a few instructional vignettes, including a step-by-step guide for writing the scoring functions, guidelines for writing the instrument descriptions, and templates for writing the function documentation.
Make the unit testing framework of PROscorer and PROscorerTools more comprehensive. Most of the code underlying the 8 functions will be already be tested by the PROscorerTools tests; however, I intend to come up with a standard set of tests for PROscorer functions to make it easier for me and others to add unit tests to their scoring functions.
Write some educational vignettes on PRO scoring methods and best practices.
Add capability to generate IRT-based scores for PROs that use that scoring method. I know many researchers that use various PROMIS measures. They would prefer to use the IRT-based scoring method, but find it too difficult to integrate into their research workflow. PROscorer could make IRT-based scores accessible to a much wider group of researchers.
You can access the “help” page for the PROscorer
package by typing ?PROscorer into R.
Supplemental documentation, including the Instrument Descriptions, can be found in the PROscorer vignettes. You can access them from the main PROscorer webpage on CRAN. More vignettes will be added in future updates.
If you need to score PRO measures from the the FACT (Functional Assessment of Cancer Therapy) and FACIT (Functional Assessment of Chronic Illness Therapy) family of measures, please see the FACTscorer package.
The underlying code base of the PROscorer package is built from functions from the PROscorerTools package.
The PROscorerTools documentation and vignettes might also be helpful. You can access them from the main PROscorerTools webpage on CRAN. More vignettes will be added in future updates.
If you have any feature requests, or you want to report bugs or other strange behavior in PROscorer, please submit them to me on the PROscorer GitHub page.