One of the objectives of the study was to see whether clinicians would accept the PAR since it was introduced by an external group with no specific incentives and without close supervision. We found that the PAR was in general well accepted and at the end of the pilot study 84% of the children were being admitted using the PAR reflecting a gradual improvement over time plausibly related to both the ongoing training and the 2 monthly, open feedback sessions (Table ).
There was, however, reasonable variation (53% to 100%) in use of the PAR between clinicians (Figure ). Factors associated with lower use of the PAR included later training, admission with a non-IMCI diagnosis (e.g. burns or a surgical condition) and clinical cadre. The association with later training may be confounded by a more general problem of poor motivation. Those who delay attending training may be least likely to act on any new initiative. More active supervision might be expected to address this problem. Thus, in a study by Duggan et al
, observation of primary care providers was found to correlate positively with improved quality of records [2
]. Nurses (NO) were in general responsible for admitting newborn babies to the nursery and had a lower rate of use of the PAR (figure ). Informal comments made suggested that some nurses felt uncomfortable committing themselves to writing the 'medical record', seeing the use of the PAR as the clinician's duty even if the clinician was not expected to see the child until the next day.
There was a marked improvement in documentation over time for most specific disease related signs such as conscious level, chest indrawing, and skin turgor (table ). As documentation did not improve for these features for children admitted in the post-intervention period when a PAR was not used the improvement can be attributed to the PAR and is in keeping with several studies showing that structured admission forms result in marked improvements in data recording [5
]. The fact that the recorded results of clinical symptoms and signs assessed were predictive of death argues that the PARs were completed meaningfully and that the improved data was of value. However, data on admission weight, immunisation status and respiratory rate were each missing from 40% of records or more even though the data were clearly required in the PAR.
Why does performance vary markedly for individual symptoms and signs? Our study was not designed to address this question. However, we can speculate that several factors may contribute. The majority of symptoms and signs can be evaluated rapidly and the PAR permitted rapid documentation (circling an option). A minimum of effort was therefore required and in fact, informally, many clinicians commented that the PAR made admitting patients quicker, again consistent with work showing that a well designed standardized admission form could reduce the workload of the admission team and length of admission clerking[10
]. Counting the respiratory rate or sending the patient out to be weighed, however, require somewhat more effort which alone might result in a failure of documentation. In addition, perhaps these data and that of immunisation status might be regarded as more of a nursing process than a medical process making the clinician happy to ignore them. Failure to record these data may also reflect a perception that they are not important (with which we disagree), or at least not important enough to warrant the additional effort required. Whatever the reason it is possible that a more active process of supervision might address these failings.
Although our results are encouraging there are clearly potential issues of sustainability. Introducing a new inpatient form is associated with some costs. However, the fact that the form is a single sheet of paper used for inpatients alone meant that producing the each form cost only 3 – 4 Kenyan shillings or approximately US$ 50 per 1,000 admissions, even if sustained using a standard photocopy. Costs could be considerably lower if forms were commercially printed. Cost is not the only issue in sustainability. We do not know whether high utilisation rates of the PAR were sustained after the withdrawal of feedback and it may be that some supervision or oversight of the process is required on a continuous basis.
A limitation of this study was that it was done in only one district hospital making it hard to generalise about likely performance elsewhere even though the results are very encouraging. Furthermore improving documentation, while of some value in its own right as an information tool, may have no direct effect on the quality of care offered to children, a far more important outcome. However, as an aide to ensuring a comprehensive clinical review and promoting decisions on disease severity based on evidence the PAR may be a useful first step to improve the care of children admitted to hospital with conditions responsible for most deaths. One disadvantage of a standardised approach, however, is the danger that the resultant focus on specific tasks may limit the breadth and depth of any one child's evaluation.