Skip to contents

Compute CIHI Hospital Frailty Risk Score

frailty_score.Rd

frailty_score returns the number of frailty conditions for hospital admissions based on the CIHI hospital frailty risk score (HFRS). By setting component_wise to TRUE, function alternatively returns identified frailty-related diagnosis codes and their corresponding frailty conditions.

Usage

frailty_score(cohort, ipdiag, erdiag, component_wise = FALSE)

Arguments

cohort

(data.table, data.frame)
Cohort table with encounters of interest and their corresponding age. Each row corresponds to a single encounter. Must contain GEMINI encounter ID (genc_id), and age of the encounter (age).

ipdiag

(data.table, data.frame)
ipdiagnosis table as defined in the GEMINI Data Repository Dictionary. This table must contain the genc_id and diagnosis_code fields in long format. The diagnosis codes must be free from any punctuation or special characters.

erdiag

(data.table, data.frame)
erdiagnosis table as defined in the GEMINI Data Repository Dictionary. This table must contain the genc_id and er_diagnosis_code fields in long format. The diagnosis codes must be free from punctuation or special characters.

component_wise

(logical)
Default is FALSE. When TRUE, the function does not aggregate the score and instead outputs a table in long format that shows the frailty conditions contributed to the score for each encounter. Encounters with no frailty conditions found (frailty_score_derived=0) are excluded.

Value

(data.frame)
If component_wise == FALSE: genc_id (numeric),
frailty_score_derived (numeric) total number of frailty conditions for the encounter.

If component_wise == TRUE: genc_id (numeric),
diagnosis_code (character),
ICD-10-CA diagnosis codes associated with the encounter frailty_categories (character)
frailty category of the ICD-10-CA code mapped to

Details

The CIHI HFRS is a contextual measure of frailty for patients aged 65 years and older. It categorizes a list of ICD-10-CA diagnosis codes into 36 distinct frailty-related conditions. The level of frailty is determined as the cumulative number of distinct frailty conditions (an equal-weight algorithm) present in an individual. ALL diagnoses are included to the calculation (i.e. all emergency department (ER) and in-patient (IP) diagnoses, all diagnosis types). When none of the frailty conditions is present in an individual, a score of zero is assigned.

The function closely adheres to the CIHI HFRS with the following adaptations:

  • No look-back period: Score is computed at encounter level. The 2-year look-back in the CIHI HFRS is not implemented. This adaptation systematically underestimates frailty but ensures comparable scores across time and hospitals considering variations in data availability

  • Score format: Integer scores are returned representing the sum of the number of frailty conditions. These scores can be easily converted to the different formats (i.e. continuous fractions, 8 risk groups, binary) defined by CIHI HFRS. For example, dividing the returned score by 36 (maximum number of conditions possible) gives the continuous CIHI HFRS. Users interested in further categorizing the scores should refer to Amuah et al, 2023.

Warnings

A warning message is returned when the user-provided cohort includes encounters not qualified for assessing frailty. Encounters below the age cutoff for frailty assessment are excluded from the results. Encounters with no diagnosis data are excluded from the results. The size of exclusion is detailed in the warning message. frailty_score_derived for these encounters are NAs, when joining the result table to the original cohort.

A warning message is returned if erdiag is explicitly set to NULL. It's possible to exclude diagnoses at the emergency department from the consideration by explicitly specifying erdiag=NULL. However, excluding any types of diagnoses is not recommended. The CIHI frailty score was developed and validated based on all diagnoses present in medical records (both NACRS and DAD). Excluding diagnoses in NACRS was found to underestimate frailty levels (Amuah et al, 2023).

Notes

The original development paper of CIHI HFRS maped 595 ICD-10-CA diagnosis codes to frailty conditions. Three codes (Z96.62, U07.1 and U07.2) were added to this mapping in the CIHI methodology notes. This function uses the mapping of 598 ICD-10-CA diagnosis codes in the CIHI methodology notes to identify frailty conditions.

The previous frailty_score() function calculated the UK HFRS (Gilbert, 2018), and it is now deprecated. Using a similar approach as the UK HFRS, the CIHI HFRS was developed and validated based on Canadian cohorts, making it particularly suited for GEMINI data. The UK version remains available in Rgemini version 0.3.1 and earlier but will not receive future maintenance. Users interested in the UK version should refer to the original publications for important differences in diagnostic coding practices and age threshold.

References

We recommend referencing both original articles (UK HFRS and CIHI HFRS) when using this function:

  • UK HFRS: Gilbert T, et al. Lancet, 2018. http://dx.doi.org/10.1016/S0140-6736(18)30668-8

  • CIHI HFRS: Amuah JE, et al. CMAJ, 2023. https://doi.org/10.1503/cmaj.220926

  • CIHI methodology notes

Examples

if (FALSE) { # \dontrun{
cohort_dum <- data.table(genc_id=c(1, 2, 3), age=c(64, 65, 80))
ipdiag_dum <- dummy_diag(nid=3, nrow=10, ipdiagnosis=TRUE, pattern = "C20$|R460$") # frailty conditions
erdiag_dum <- dummy_diag(nid=3, nrow=5, ipdiagnosis=FALSE, pattern = "M121$") # not a frailty condition
frailty_score(cohort_dum, ipdiag_dum, erdiag_dum, component_wise = FALSE)
} # }