Format

Send to

Choose Destination
Med Decis Making. 2015 Nov;35(8):959-66. doi: 10.1177/0272989X15598540. Epub 2015 Aug 5.

The CREATE Method for Expressing Continuous Outcome Data in Absolute Terms for Use in Patient Treatment Decision Aids: A Validation Study.

Author information

1
Faculty of Health Sciences, McMaster University, ON, Canada, (MM, SLC, HMA)
2
University of Toronto, Toronto, ON, Canada (JCV, KM, SOM, JGH, PC, BS)
3
University Health Network, Toronto, ON, Canada (NJ)
4
Hamilton Health Sciences, General Division, Hamilton, ON, Canada (RM)
5
Montreal Heart Institute, Montréal, Québec, Canada (EMJ)
6
Libin Cardiovascular Institute of Alberta, University of Calgary, Calgary, AB, Canada (JS)
7
Ottawa Cardiovascular Centre, Bank Street Professional Centre, Ottawa, Canada (JN)
8
University of Ottawa Heart Institute, Ottawa, ON, Canada (RB)
9
University of British Columbia Island Medical Program, University of Victoria, Victoria, BC, Canada (NS)
10
University of Ottawa and Ottawa Hospital Research Institute, Ottawa, ON, Canada (DS).

Abstract

BACKGROUND:

Patient decision aids (PtDAs) supplement advice from health care professionals by communicating the absolute risk or benefit of treatment options (i.e., X/100). As such, PtDAs have been amenable to binary outcomes only. We aimed to develop and test the validity of the Conversion to Risk Estimates through Application of Normal Theory (CREATE) method for estimating absolute risk based on continuous outcome data.

METHODS:

CREATE is designed to derive an estimate of the proportion of those who experience a clinically relevant degree of change (CRDoC). We used a 2-stage validation process using real and simulated change score data, respectively. First, using raw data from published intervention trials, we calculated the proportion of patients with a CRDoC and compared that with our CREATE-derived estimate using chi-square tests of association. Second, 200,000 simulated distributions of change scores were generated with widely varying distribution characteristics. Actual and CREATE-derived estimates were compared for each simulated distribution, and relative differences were summarized graphically.

RESULTS:

The absolute difference between the estimated and actual CRDoC did not exceed 5% for any of the samples based on real data. Applying the CREATE method to 200,000 simulated scenarios demonstrated that the CREATE method should be avoided for outcomes where the underlying distribution can be reasonably assumed to have high levels of skew or kurtosis.

CONCLUSION:

Our results suggest that standard statistical theory can be used to estimate continuous outcomes in absolute terms with reasonable accuracy for use in PtDAs; caution is advised if outcome summary statistics are assumed to have been derived from highly skewed distributions.

KEYWORDS:

continuous outcomes; normal statistical theory; risk communication; validation study

PMID:
26246516
DOI:
10.1177/0272989X15598540
[Indexed for MEDLINE]

Supplemental Content

Full text links

Icon for Atypon
Loading ...
Support Center