Are needs for new cardiovascular prediction rules clearly assessed in derivation?: a review of derivation studies included in the International Register of Clinical Prediction Rules for Primary Care
Before any new research is undertaken, a careful review of current knowledge on a particular subject must be carried out to identify pertinent research questions, choose optimal study designs and avoid unnecessary duplications. Although an increasing number of new cardiovascular prediction rules are produced, it is not known whether their needs are clearly assessed by examining existing prediction rules. The primary objective of this study was to examine whether authors of derivation studies justified new cardiovascular prediction rules by examining existing prediction rules. The secondary objective was to understand why it was necessary to develop a new cardiovascular prediction rule when prediction rules for the same problem already exist.
Derivation studies of clinical prediction rules included in the cardiovascular domain of the International Register of Clinical Prediction Rules for Primary Care were evaluated. The introduction section of each derivation study was examined to identify citations for existing cardiovascular prediction rules. A content analysis explored the reasons for determining the existing prediction rules were insufficient and therefore, a new prediction rule was necessary for the cardiovascular problem.
The largest numbers of prediction rules were available for Stroke/cerebrovascular accident (K90, n = 29), Risk factor cardiovascular disease (K22, n = 26) and Pulmonary embolism (K16, n = 16). Of 85 derivation studies that met the inclusion criteria, 44 derivation studies (51.8%) cited existing prediction rules, 35 derivation studies (41.2%) did not cite any existing prediction rule, and 6 derivation studies (7.1%) declared there was no prediction rule to cite. Fifty derivation studies (58.8%) were conducted in the United Kingdom and the United States combined which included 218447 participants (87.1% of the participants in all studies). In a thematic content analysis of derivation studies that citied existing prediction rules, 6 categories for insufficiency of existing prediction rules were generated: (1) derivation related (e.g. selection bias in derivation study), (2) construct related (e.g. inappropriate outcome or missing important predictor variable), (3) performance related (e.g. poor sensitivity or specificity), (4) transferability related (e.g. did not perform well in external validation), (5) evidence related (e.g. lack of external validation), and (6) simple citation (e.g. citation of existing rule without providing any insufficiency).
New cardiovascular prediction rules are often derived without considering existing prediction rules which may have led to creating many duplicate prediction rules for the same cardiovascular problem. Producing an entirely new prediction rule may not be the most efficient way to address some of the insufficiencies (e.g. lack of external validation) of existing cardiovascular prediction rules described in derivation studies.