Reliability, completeness and accuracy of cleft subphenotyping as recorded on the CLEFTSiS (Cleft Service in Scotland) electronic patient record

ObjectivesThe CLEFTSiS (Cleft Service in Scotland) electronic patient record (EPR) contains pre-surgical photographs among other records and this study proposes to assess the reliability of diagnosis and accuracy of the type and extent of cleft.Design and settingRetrospective study, University of Dundee.Materials and methodsFrom all 1092 cases contained by the CLEFTSiS EPR from inception (April 2000) until September 2011; syndromic, atypical and submucous clefts, still births, abortuses, cases with non-cleft velopharyngeal incompetence, Pierre Robin Sequence, or Simonart's bands were excluded to leave 730 case records for analysis. Observer 1 recorded the LAHSAL subphenotype using the pre-surgical photographs for each case. Images that were absent, only partially demonstrated the intended anatomical site or were unclear were excluded. The data for Observer 1 was compared to the EPR subphenotype using the Kappa statistic. Thirty randomly selected cases were recoded one month later to calculate intra-observer reliability and three further observers subphenotyped these cases to determine inter-examiner reliability, both using Kappa statistics.ResultsIntra- and inter-observer reliability of the LAHSAL system were excellent (0.809–0.992). The correspondence between Observer 1 and the original CLEFTSiS subphenotype was almost perfect (0.812–0.862) except the soft palate and the left alveolus, where agreement was substantial (0.638–0.776). Only 68 (9.3%) of the records fully demonstrated all six anatomical areas of the LAHSAL coding system with intra-oral images being the most frequently deficient views.ConclusionsThe subphenotype data held on the CLEFTSiS EPR is accurate but incomplete. Only 9.3% of cases contained all relevant pre-surgical photographs. Subphenotyping using the LAHSAL classification has a high degree of intra- and inter-observer reliability.