An international call for validation studies of the C-Model
was issued in February 2014. Members of the WHO MCS
research network, the International Network of Obstetric
Surveys Systems (INOSS), and independent researchers
received an invitation to test the C-Model in their databases.
To ensure standardisation, an external validation protocol
with detailed instructions including variable recoding, computing probabilities of CS using the C-Model coefficients,
generating ROC curves, and other information were provided. The researchers that responded to the call returned
their results via e-mail and the information was compiled in
a master spreadsheet (Supporting Information Appendix
S2). For each database, estimates generated by C-Model versions with incomplete data (i.e. when a variable was not
available in the dataset) were discarded. Random effects
meta-analyses were used to generate summary estimates of
areas under the ROC curves. Based on the results of these
meta-analyses, the C-Model version with largest number of
variables was considered able to generate the ‘best estimate’
of CS probability. The deviation of the observed CS rate (and
an uncertainty range) was determined for each database considering the best estimate of CS probability. This uncertainty
range was arbitrarily defined by the authors as 20% because
differences >20–25% are commonly considered clinically significant or appreciable differences.24 The ratio between
observed and predicted CS (the standardised caesarean
section ratio) was determined
An international call for validation studies of the C-Modelwas issued in February 2014. Members of the WHO MCSresearch network, the International Network of ObstetricSurveys Systems (INOSS), and independent researchersreceived an invitation to test the C-Model in their databases.To ensure standardisation, an external validation protocolwith detailed instructions including variable recoding, computing probabilities of CS using the C-Model coefficients,generating ROC curves, and other information were provided. The researchers that responded to the call returnedtheir results via e-mail and the information was compiled ina master spreadsheet (Supporting Information AppendixS2). For each database, estimates generated by C-Model versions with incomplete data (i.e. when a variable was notavailable in the dataset) were discarded. Random effectsmeta-analyses were used to generate summary estimates ofareas under the ROC curves. Based on the results of thesemeta-analyses, the C-Model version with largest number ofvariables was considered able to generate the ‘best estimate’of CS probability. The deviation of the observed CS rate (andan uncertainty range) was determined for each database considering the best estimate of CS probability. This uncertaintyrange was arbitrarily defined by the authors as 20% becausedifferences >20–25% are commonly considered clinically significant or appreciable differences.24 The ratio betweenobserved and predicted CS (the standardised caesareansection ratio) was determined
การแปล กรุณารอสักครู่..
