MJA: Reviews of Porteous, et al., Management of childhood gastroenteritis in the community

Reviews of

Management of childhood gastroenteritis in the community

Perceptions of general practitioners

Jennifer E Porteous, Richard L Henry, Edward V O'Loughlin, Malcolm Ireland, J Lynn Francis and Robyn G Hankin

Reviewer 1: Dr Mark Ferson, Director, Public Health Unit, South Eastern Sydney Area Health Service

I found this manuscript, which describes a survey of GPs' self-reported practice in the management of acute gastroenteritis in young children, to be of interest. This is a common condition, and virtually all affected children would present to a GP at a stage where deterioration requiring hospitalisation could be prevented by appropriate management advice, particularly regarding the use of oral rehydration solution or equivalent.

My comments follow:I believe that the study design is satisfactory. In Methods, it would be useful to mention the time of the year when the survey was conducted, as gastroenteritis in young children has a strong seasonal component related to rotavirus circulation. This should also be touched on in the Results under "Prevalence of . . ." and the first paragraph of the Discussion.

[Authors' response: The time of year when the study was conducted is now mentioned. The prevalence of gastroenteritis has now been omitted in the interests of shortening the article, as requested by the Journal's editorial committee.]

Reference to Australian guidelines for management of acute gastroenteritis such as the Gut Foundation's Diarrhoea in children, if these were used to judge the adequacy of reported management, would be valuable in the Analyses section of Methods.

[Authors' response: Reference to the basis of optimal management is included in the Analysis section of the Methods and a box summarising currently recommended optimum management is enclosed.]

Results: on p10 in the para starting "Table 2 . . ." the 3rd sentence becomes a little confusing and appears to need a little more plain English. Perhaps dropping "of providing optimal advice" from the end of the sentence may help. I would defer to a statistician for a more expert examination of this paragraph and the accompanying Table 2.

[Authors' response: The section on logistic regression has been completely rewritten and Table 2 has now been deleted. The new Table 2 was Table 3.]

Discussion : on p13, paragraphs 2 and 3, comparisons with the authors' earlier study [7] needs to be explained further by them. The present study is a survey of GPs whilst the earlier study was of pre-admission management of hospitalised children, a small and selected subset of children seen by GPs for gastroenteritis. This difference needs to be acknowldeged, especially in para 3 line 3 ". . .in the same population less than . . ." (a better word may be community or region , rather than population which has a specific technical meaning in epidemiology).

[Authors' response: The Discussion now highlights the fact that this is a survey of GPs while the earlier study was of preadmission management of hospitalised children. The word "population" no longer appears so that any potentially misleading inference has been removed.]

The concluding paragraph (p16) line 6 (and the abstract) uses the word structural , jargon which may be replaced by a more meaningful term (perhaps practical, although this is not really adequate either). On the next line, paternal is the victim of reliance on the ubiquitous spell checker (!), and should be parental, I imagine.

[Authors' response: The word "structural" has been removed, and "parental" corrected.]

It may be worth summarising in a box what is currently considered optimal management of childhood gastroenteritis.

[Authors' response: Done.]

Table 1 suffers from a surfeit of abbreviations (is there a collective noun for abbreviations?), particularly those in the "variable" column at left, and the Kwallis in the top right, which is short for Kruskal-Wallis statistic. Table 2 also needs a little attention in this respect.

[Authors' response: Table 1 has been rewritten to avoid abbreviations. Table 2 has been deleted.]

The abstract is well written, apart from my objection expressed earlier to the word structural.

Reviewer 2: Professor Frank Oberklaid, Centre for Community Child Health and Ambulatory Paediatrics, Royal Children's Hospital, Melbourne, VIC

The authors are correct in asserting that gastroenteritis is an important cause of paediatric morbidity, and provide a very good rationale for the study together with a good literature review. It is pleasing to see academics working in tertiary care institutions begin to look beyond just treating the patient and attempting to improve management in community settings. Studies like this are to be welcomed.

While the response rate for a study of this type (69%) is probably satisfactory, I am interested in those who have not responded. While the authors draw attention to the possibility of selection bias, it would be nice to provide data on non-responders. I presume from the way this study was conducted that data may not be available - if this is the case then the authors should state this in the results.

[Authors' response: The comment (no information available on non-responders) is now included.]

While the methodology of the study is reasonable (a case history requiring narrative comment) this results inevitably in "clumping" results. It would have been interesting also to have some information in a more quantitative way, e.g., respondents' level of knowledge of issues involved in assessment and management of children with gastroenteritis. It would be interesting to see some sort of discussion around the limitations of this sort of methodology, which on the one hand may provide some rich data but on the other may miss out on quantitative data.

[Authors' response: Presenting some of theinformation in a more quantitative way would have required a lot more space than was available for the paper.]

The authors draw attention to one of the limitations of this methodology, as described above, in that it may not really be a test of knowledge but rather the respondents may have given "socially desirable" answers. This is methodological problem of all such studies, and indeed of any educational intervention, in that in it is difficult to specifically demonstrate that there is a change in actual practice as opposed to improvement in knowledge. The authors may wish to comment on this difficulty.

[Authors' response: The issue of respondents giving "socially desirable" answers is now specifically mentioned, together with the inherent difficulty of such studies.]

In Table 1 it would be clearer if the strategy was actually spelled out, rather than have abbreviations and then explanations as footnotes.

[Authors' response: Done.]

I believe that Table 2 could easily be omitted as it is well reported in the text.

[Authors' response: As suggested, Table 2 has been deleted. The new Table 2 was Table 3 ]

Statistical Reviewer: Dr Michael Adena, INSTAT Australia, Pty Ltd, Canberra, ACT

This paper competently summarises a mail survey of GPs in the Hunter region of NSW. The survey instrument appears suitable, the participation rate of GPs is satisfactory for this type of surveyand the description of the survey and its analysis are comprehensive and well presented.

Some minor queries relate to the statistical material:Some percentages are not correct. For example, the first percentage in the Results in the Abstract, quoted as 69%, should be 68%. In several places, the expression "almost x%" is used, but the exact percentage, which could be quoted in most instances without requiring the spelling out of the number, is one or two percentage points less.

[Authors' response: The incorrect percentages have been addressed. Terms like "almost x%" have been deleted.]

In general, the number of responses reconcile to totals or subtotals. One exception is that there are only 287 respondents accounted for (instead of 289) in the first seven lines of the Results in the Abstract.

[Authors' response: 289 respondents are now accounted for.]

The degrees of freedom for the Kruskall-Wallis test statistic should be quoted in the text and in Table 1. In the second paragraph of "Assessment of management strategies", I assume that the optimal advice percentages are quoted but that the test statistic was calculated using 3 categories of advice. If the female/male comparison were also to use 3 categories (excluding the "vague advice" category) I suspect the difference would be statistically significant.

[Authors' response: Degrees of freedom are now quoted. The optimal advice percentages are now presented in a clearer manner and female/male comparisons have been analysed as suggested - and the difference is statistically significant.]

The reporting of logistic regression is incorrect. I assume that the reason why the odds ratio for rural vs urban reported in the text differs from that in Table 2 is that the former controls only for age, while the latter also controls for sex and the number of patients seen weekly with vomiting and diarrhoea. If so, it is curious that the odds ratio for age does not also differ slightly.

[Authors' response: The reporting of logistic regression was incorrect and has now been presented in percentage form as suggested.]

Odds ratios are not multipliers for percentages, as has been assumed in "Assessment of management strategies". For example, the odds ratio of 4.17 for rural vs urban might reflect that 85% of rural and only 54% of urban GPs offer optimal advice - the difference is about 30 percentage points, not "four times" as claimed in the text. Similarly, for each decade of age, the decrease in the percentage of GPs offering optimal advice is about 10 percentage points, not "about two thirds". (Incidentally, an upper bound for the difference in percentage points is 25 ln OR, and the difference decreases as the two percentages being compared differ from 50%.)