Hi All
I need your guidance on how we can combine multiple variables and develop one index.
The goal is to have a index based of which we can plan for staffing. The variables are: daily patient volume,
%admitted,
% severity.
For example if the daily vol is high, also % admitted is high, and % severity is also high then we give more score lets say 3.5 which means we have to plan for more nurses vs if vol is high but severity is low means we might not need more nurse even the vol is high because the severity is low ( low severity = less attention needed)
data lay out:
Date Patient daily vol % admitted %high severity
10/20 250 35% 10%
Any guidance will be highly appreciated.
Thank you
Hello Kamu,
The answer depends on what you want to know, and how the index score would be used. Please note that principal component analysis, if applied to your three scores, would yield some composite that may or may not be helpful for your specific purpose.
If you wanted to know, what linear combination of the scores best corresponds with some existing criterion variable (e.g., bed usage), then multiple regression would likely be most useful.
If you wanted to know, how medical professionals would rate the seriousness of any specific collection of the three variables (for example, as in a triage situation), then you'd likely fare better by having a group of experts review various profiles, and determine which combinations yield non-ignorable differences in perception of seriousness. Do know, however, that in many applications, clinical judgment fares worse than actuarial/statistical judgment as regards any number of outcomes.
Apologies if I've misunderstood your query, perhaps some elaboration about how the resultant composite score is to be used would help to elicit constructive advice.
Good luck with your work.
Thank you so much David for your response.
Let's keep PCA aside.
Our goal is to staff ( nurses) as per the composition of patient ( meaning what level of care required /resources ) likely to arrive on that particular day. The elements that we are considering to determine the scores are patient volume( higher vol required high staff/resources), 2.patient severity ( trauma cases required high staff/resources) and 3. % admission ( high % admission required high staff/resources).
how should I proceed?
Thank you
Principal component analysis (PCA) as a pattern recognition technique will help you to know and ascertain which parameter among your variables is more important than others. Rather than being subjective in assigning weightage to your selected variables, PCA will help you to justifiability assign the weightage based on their variability.
There are about four steps you may follow in developing an index. You can refer to my published paper on index development of water quality.
Refer to my steps in developing the index in the link below, and relate it to your research work. PCA was used in the parameter selection processes, and the weightage assigned can be justified using PCA.
http://www.academia.edu/download/58389790/Development_of_water_quality_index_of_ex-mining_ponds_in_Malaysia.pdf
Best of luck.
Thank you Isa. I will go over your paper and get back to you if more questions.
Hello again Kamu,
From your follow-up post, it sounds to me as if you'd like to be able to predict the subsequent staffing required (the outcome variable), based on the three variables you'd like to use (volume, admitted, severity). I'd suggest you go back in time, note each of the three indicator variable values for a given day, and the subsequent staffing that was required. Over a large enough number of days, this would give you the data needed to develop a regression model for estimating needed staffing.
While the day to day values are likely to be correlated, and therefore "cases"won't be strictly independent, if your interest is solely with your agency/institution, this approach could, nevertheless, help to make informed decisions about staffing.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Indexes