Calculating hospital frailty risk score

[Abdel Chaar]

Hi,
I'm working on the National Inpatient Sample (NIS) and wondering if anyone can help me finding a way calculate the "Hospital Frailty Risk Score". It's validated score based on 109 ICD-10 codes where each code gets a certain score and all get sumed together as the frailty score. Is there anyway we can do this similar to charlson comorbidity index in STATA?

Here few couple examples of studies utilizing this score in NIS:
https://pubmed.ncbi.nlm.nih.gov/33632638/
https://pubmed.ncbi.nlm.nih.gov/34982727/

Thank you

[Weiwen Ng]

The HFRS isn't a commonly used instrument, and there's no written Stata program to calculate it. Now, you have a couple of options. First, the authors in the second link used Stata, so you could email the lead author to ask for the code. However, academics often get a lot of emails, and the lead author may not be the coder, and they may be unable to get in touch with the person who did the coding. You can feel free to email a few times to remind them.

If you trace that study's citations, they cite numbers 13 and 14 as the background for the HFRS. Citation 13 leads to this Lancet article. If I read the article correctly, table A2 of their appendix is the list of the ICD-10 codes that go into the HFRS algorithm and how many points are awarded for each. The HFRS score is just the sum of every observation's total score. The codes are actually ranges, e.g. they say F00, Dementia in Alzheimer's Disease, is worth 7.1 points, but they really mean all ICD-10 codes beginning with F00, which I believe is a total of 5 codes. You should be aware of the ICD10 suite of commands, which (among other things) let you generate an indicator variable by specifying a range of codes or all codes beginning with a certain prefix, the latter of which should enable you to code the HFRS.

[Leonardo Guizzetti]

Weiwen's suggestions are very helpful. Also be aware that Stata contains two ICD10 commands. -icd10- is for the WHO set of codes, and. -icd10cm- is the US-specific version. This may matter for the source of your dataset and score creation.