Towards Comprehensive and Accurate Recording of Cerebral Palsy Diagnoses in Registries: Challenges and Implications

Postneonatal cerebral palsy (CP) differs from pre- and peri-neonatal CP in the timing (beyond the neonatal period up to 2 years of age) and the nature of the causal event leading to brain injury. Population-based registers usually monitor this relatively rare type of CP [1]. Their general objective, from a public health perspective, is to provide reliable estimates of prevalence over time [2]. Unlike hospital-based registers, population-based registers cover a broader spectrum, encompassing entire populations and providing a more accurate picture of trends, risk factors, origins, and burden [3]. For postneonatal CP, such registers have reported a decline in prevalence in high-income countries [4]. However, one of the challenges is also to reliably document the evolution of the causes [5] (which are most often well identified), to assess the effectiveness of preventive health strategies, possibly. The article by Waight and colleagues [6] reports on the impact of a change in the operating procedures of their CP register in Australia about the surveillance of postneonatal CP. The authors describe using a new data source for case inclusion in the register. Due to the specificity of the overall care pathway for postneonatal CP in their surveillance area, this is an essential addition to the existing routines. The authors also report new procedures for identifying postneonatal CP cases that are not always clearly identified as CP in the medical records. These new procedures result in a substantial change in prevalence. This publication highlights and reminds registers of the importance, but also the difficulty, of striving for completeness in collecting clinically validated data for postneonatal CP. Of course, as postneonatal CP is rare, the impact of these new inclusions on prevalence is even more significant. However, all these considerations should be taken much more broadly and serve as a reminder of the exemplary operating practices that should be standard in any register. Indeed, for public health practice and epidemiological research, population-based registers are the gold standard for providing information on CP. They intend to identify cases and collect information from multiple sources on all individuals diagnosed with CP within a clearly defined population, for which general population data are available and serve as the denominator for calculating rates. The first comment therefore concerns the completeness of registration, especially the impact of the multiplicity of data sources used for case identification and inclusion. A clear distinction should be made between the data sources used to identify cases of CP that meet the inclusion criteria of the register, and data sources that can provide additional information on children already included. Using multiple, complementary, and independent data sources to detect and identify cases of CP is essential to guarantee the objective of completeness and therefore the accuracy of the prevalence, as has already been clearly demonstrated in a Norwegian study by Hollung et al. [7] The basis of the work is, of course, to cover mainly medical data sources, either directly by reporting or recording cases by the diagnosing doctors, or indirectly by consulting medical records. However, it is also necessary to consult other data sources, such as medical-administrative databases and/or school or special education databases, but also data sources that depend on the care pathways that can be specific to a surveillance area, a subpopulation, or a pathology (as is the case here for postneonatal CP), which are sometimes less easy to identify. The omission of an important data source will therefore obviously affect prevalence. However, if the omitted data source is associated with a particular care pathway, sub-population, or pathology, this will also impact the recorded CP profile. It is possible that the children treated in the rehabilitation services described in the article by Waight et al. [6], who were not previously included in the register, have characteristics (specific causes or severity) that distinguish them from the cases included through the usual procedures. It would therefore have been interesting to describe the causes of the brain lesions and the clinical characteristics of these children who were not initially identified. This would have made it possible to assess whether and how these additional cases might have changed the distribution of causes of postneonatal CP and/or the clinical characteristics compared to other international publications [4]. Thus, in addition to a systematic underestimation of prevalence, the omission of some data sources could potentially bias the characteristics of the population of children included. The second comment concerns the accuracy of the record and the methods used to ensure the correct identification of cases in the different data sources. Registration must, of course, be based on a validated medical diagnosis using a precise definition of postneonatal CP, and this approach should be applied more generally to all CP. As the definition of CP is based on phenomenology (clinical picture and history) and not on aetiology, it is essential to consider all cases whose overall description is consistent with a diagnosis of CP, even if the exact wording of CP is not used as the final diagnosis in the medical record or in the data source used to identify the case. There may be a particular issue with postneonatal CP, where the exact wording of CP may be used less frequently in clinical practice than other forms. However, this consideration and the use of a decision tree, as presented in the article by Waight et al. [6], should be a prerequisite for registering all cases of CP. A very similar decision tree was initially proposed as a common language and minimum harmonisation for including CP cases in all registries of the Surveillance of Cerebral Palsy in Europe (SCPE) network [8]. This decision tree also applies to postneonatal cases, as neither the definition nor the inclusion in the register depends on the cause. In addition to the clinical description, only the time of occurrence of the lesion is required to determine its postneonatal origin. It seems essential that active registration, i.e., identification of cases by the registry medical team based on such a decision tree, is carried out routinely, possibly in addition to systematic and potentially passive reporting. Conversely, a diagnosis of CP mentioned in a data source cannot be relied upon without careful verification of the diagnosis [9], especially if the data sources are medical–administrative or school databases—the study by Hollung et al. [7] clearly showed how adding an administrative data source significantly improves comprehensiveness, but that a review of the entire hospital record of the children is essential to exclude cases in which a CP diagnosis code has been incorrectly assigned. The work of Waight et al. [6] is a very relevant and helpful reminder that running a register is a live process. Ideally, registers should regularly review their data to ensure they are complete, accurate and unbiased, to maximise the potential impact of the information provided. For any work based on register data, it is essential to present a description of the data sources, how data are collected and the methods used to ensure the validity of data, in addition to key elements such as design, surveillance area, definition of the study population and eligibility criteria. If it is suspected that the prevalence may be affected by a lack of completeness, it is important to quantify and above all qualify ‘who’ the omitted cases are, in order to determine whether there is a risk of systematic bias in the type, origin, cause and/or severity of CP. Finally, the very simple but highly ambitious issues of a CP register are as follows: Where are the children with CP? And how can CP diagnoses be correctly and accurately identified in the different data sources? The author takes full responsibility for this article. The authors have nothing to report. The authors declare no conflicts of interest. The authors have nothing to report.