Articles reporting descriptive epidemiological studies have become passé in the injury prevention literature. What we all crave is reports of interventions that make a difference.1 Yet, the number of articles reporting descriptive studies submitted for publication continues to exceed that of articles reporting intervention studies. The publication of three such articles in this edition of JSAMS offers an opportunity to reflect on the conduct and reporting of the lowly descriptive study.
Most epidemiological investigations begin with a hunch that a problem exists. The first step in investigating such hunches is to estimate the size of the problem and describe its basic characteristics; that is, to undertake a descriptive study. Typically, researchers will seek existing data sources that they can use for this purpose; commonly, hospital inpatient data. Mitchell and Hayen 2 describe a typical inpatient data base, the Inpatient Statistics Collection (ISC), which collates data from public and private hospitals in New South Wales, Australia.
Mitchell and Hayen's aim was to demonstrate the utility of the ISC for investigating sports injury problems. For this purpose they chose a common cause of sports injury: being struck by other persons or objects, or striking against other persons or objects. Using the ICD-10 3 coded data in the ISC, they identified eligible cases, estimated the size of the problem – reporting population-based incidence rates – and described the resulting injuries by body site, diagnosis, objects involved, place of occurrence and activity. One strength of the ISC is that it uses the Australian modification of ICD-10 (ICD-10-AM), which in its more recent editions has codes for more than 260 sport and recreational activities.4 In discussing their findings, Mitchell and Hayen identified several limitations of the ISC data, particularly the lack of detailed information on the circumstances of injury. They suggest that the addition of narratives would enhance the utility of the ISC for identifying injury prevention options.
Happily for Conn, Annest, Bossarte et al., authors of the second article in this series,5 the data base they used included such narratives. The problem they investigated was that of violence in sport and recreation in the USA. Their aims were to estimate the size of this problem, determine what proportion of injuries could be attributed to violence, and describe the characteristics of violence-related sport and recreational injuries. Their source of data was the National Electronic Injury Surveillance System (NEISS), which collates data on emergency department presentations for product-related injury. Conn, Annest, Bossarte et al. used code and narrative searches to achieve their aims, provide informed comment on the size of the problem and offer sensible suggestions for its reduction. In discussing the limitations of their research, Conn, Annest, Bossarte et al. noted that there may be many more cases of violent and non-violent sport and recreational injury treated outside of hospital emergency departments.
The latter issue was of concern to Finch and Cassell, authors of the third article in this series.6 Finch and Cassell argue that incidence estimates based on hospital discharge data, such as those used in the first two studies in the series, do not capture cases of sport and recreational injury that do not make it into hospital records, but nevertheless, have a significant impact on participation. In contrast to the previous two studies, Finch and Cassell collected primary data by means of a household survey. They report population-based and participation-based incidence rates for sport and recreational injury in their study population and describe the characteristics of cases by age, gender and activity. In keeping with their aim of determining “the public health impact of injury during sports and active recreation”, they also report estimates of the impact of injury on participation and activities of daily living.
The three descriptive studies reported in this edition of JSAMS provide an opportunity to examine the advantages and disadvantages of the two types of studies involved: those using primary data collection and those using routinely collected data. Investigators using primary data collection methods can apply broad definitions of injury; collect detailed information on the circumstances and outcomes of injury; and estimate exposure-based incidence rates by collecting data on participation. On the other hand, primary data collection is costly and for this reason such studies are generally one-off.
Studies that use routinely collected data have the advantage that the data have already been collected and so avoid expensive data collection costs; they can be repeated regularly, permitting the monitoring of trends over time; and if narratives are available they have the potential to identify emerging issues (e.g., new sports). Disadvantages of secondary data sources include the definition of injury being confined to that of the source population (e.g., emergency department presentations, hospital admissions); constraints imposed on data analysis by idiosyncratic coding systems; time lags between data collection and the availability of data for research purposes; and the potential for temporal trends to be affected by changes in hospital admission policies.
If descriptive studies are an essential first step in epidemiological research, what might be done to make them more attractive to journals? First would be the standardisation of research methods, including definitions and classification systems. For example, the three studies reported in this edition of the Journal used different systems for classifying sport and recreational activities. If all researchers undertaking descriptive studies were to use the same system of classification (e.g., the activity classification in ICD-10-AM), then there would be the potential to make comparisons between studies. Second, by dispensing with lengthy discussions of putative risk factors, the length of articles could be reduced. Descriptive studies can be used to identify potential risk factors and generate hypotheses, but cohort and case-control studies are needed to test these hypotheses and confirm the significance of postulated factors. Finally, by also dispensing with recommendations for preventive measures – which not even the best-designed descriptive studies can directly support – the length of articles could be reduced even further. With the odd exception, it is premature to speculate about potential interventions from the results of descriptive studies. Just as it is necessary to confirm the significance of postulated risk and protective factors, the efficacy and effectiveness of interventions need to be investigated using appropriate research methods.
The lowly descriptive study is essential to the investigation of injury in sport and recreation, but it needs to be kept in perspective.
FULL TEXT