Results

Altogether 2281 children were triaged during the study period and 1581 of these were followed through to admission or discharge. For all children admitted final outcome was sought from inpatient records. Thirty nine per cent of children (884) were aged 1 year or younger. Forty six per cent (1042) were over 1 year but less than 5 years, and the remaining 15 per cent (350) were 5 years and over.

The majority (74%) of triage was done in the mornings. Nurses using the ETAT triage guidelines identified 92 patients requiring emergency treatment (P1) and 661 with priority signs (P2). Figure 2 shows the difference in priority allocation between the triage nurse and the assessing doctor. A total of 142 patients (6.2% of those triaged) had priorities reassigned by the assessor. Four were downgraded from P1 to P2, one of whom was admitted. Four patients were moved from P2 to P1, three of whom were admitted and one discharged. Seventy five patients were reassigned from P2 to P3, only one of whom was thought to require admission; 59 changed from P3 to P2, of whom eight were admitted.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Change in nurse allocated triage category following assessment by doctor.

Table 1 shows the agreement between triage nurse and assessor on the presence or absence of priority signs. Some signs show a much larger variation than others, but all participants learned to assess well. One of the nurses found dose calculation in the emergency treatment difficult.

Of the 1581 children triaged with follow up, 236 were admitted. Ninety per cent of P1 were admitted, 32% of the P2, and only 3.5% of the P3 patients. Altogether 85% of admissions had been prioritised to emergency or urgent categories. Thirty one children died, four while awaiting triage and 27 after admission.

During the follow up period to assess the nurses' performance when working without the supervision of an assessor, 1000 children were triaged. This assessment was carried out over five sessions; table 2outlines the results.

Sixteen children were allocated to P1, of whom 12 (75%) were admitted; 335 were felt to be P2, of whom 124 (37%) were admitted; and 649 were designated P3, of whom 17 (2.5%) were admitted.

During the study period health centre referrals did not alter in type or number, as most referrals were appropriate. It was not possible to assess whether self referrals increased with the introduction of the triage system. Mothers appeared to understand that the sickest children were being seen in advance of those who were well, and sometimes presented directly to the triage nurse if their child was in extremis.

Between 80% and 96% of patients designated P1 or P2 by the nurses during non-supervised triage were admitted. The exception is the triage conducted on the afternoon of day 4. On this afternoon there was no one available for dedicated triage duties, so the children were assessed by a trained nurse while being weighed. During this period the percentage of admissions prioritised by the nurse was only 50%.

Discussion

The study was performed to assess the possibility of training nurses and paramedical workers to triage sick children and institute life saving treatments in a developing world setting. There were large variations in allocation of emergency and priority signs. For emergency signs assessors found double the number of children with an increased capillary refill, and assessed 50% more to be cold or lethargic than the nurses did. The largest difference in allocation of priority signs was for pallor, respiratory rate, and wasting.

Of the 2281 children triaged by the nursing staff using the ETAT guidelines 142 (6%) had a different category allocated by the assessor. The category allocated seemed to correlate well with the likelihood of admission.

During the validation study in Brazil6 each patient was assessed by a nurse using the ETAT guidelines and then a paediatrician using APLS guidelines. This involved auscultation of the chest and recordings of pulse, respiratory rate, and heart rate. This was not a viable option in the Malawi arm of the study, because of the environmental limitations of space, noise levels, and large numbers of waiting patients. Therefore both nurse and assessor used the ETAT guidelines.

The two assessors in this study, although both APLS trained, came from different backgrounds, one from accident and emergency and one from paediatrics. Neither had previous experience of working in the tropics. The assessments were not blinded. The assessors were aware of nurses' allocation of priority before they recorded their own. Nurses were informed if the assessors had altered the priority, as this affected the child's placement in the department. This in itself may have affected the nurses' next few assessments.

Our study revealed large variations in the number of priority and emergency signs assessed by nurses compared to the APLS trained doctors. The correlations did not change with time but trainees screened more rapidly with practice. These variations were also much larger than those seen in the previously published study from Brazil.6 There are several possible explanations. In the Brazilian study both paediatricians responsible for assessments were from the hospital in which the trial was being run. In the Malawi arm of the study both assessors came from the UK. The variation in assessment of pallor and wasting may be caused by differences in perceived or accepted norms between western and developing countries.

Lethargy is a subjective sign, and is open to different interpretation by individuals. There is also the possibility of cultural differences and expectations regarding childhood behaviour influencing assessment.

By contrast, increased respiratory rate is an objective sign. No numerical limits had been set to define increased respiratory rate, making it impossible to ascertain the true number wrongly assessed. Much has been written about the value of respiratory rate in the diagnosis of respiratory illness in children,8-16 and its lack of predictive value for hypoxaemia in a malaria endemic area.17 There has also been considerable discussion about age specific normal values,18 the problems of accurate measurement,19-22 and its variability with fever.23 ,24 However, in this study increased respiratory rate was used as an indicator of priority rather than specific pathology. In view of the confounding factors discussed above we have not described the variations as false negatives or positives, but as differences.

The accepted gold standard for assessment of core temperature is oral or rectal thermometry. However, actual values used to define fever are often selected at random.25 In UFC the facility for rectal measurement of temperature was not available and so fever was assessed by touch. This method has been shown to have a high negative predictive value,26 ,27 and one study reported a 98% sensitivity for temperatures above 38.9°C in children under 2 years.28

It is interesting to see that the large variations in assessment of the presence of abnormal signs were not reflected in the allocation of priority categories. This may be a reflection of the experience of the nursing staff. Cioffi29 showed that triage was more likely to be correct if performed by more experienced nursing staff. The main difference was increased reference to previous patient encounters by the more experienced nurses.

It may also be that nursing staff allocated the correct priority to patients on the basis of a single sign, failing to pick up other indicators of priority. Does this mean that nurses are allocating priority on a quick eyeball assessment? The data collected at the follow up study of unsupervised triage would suggest not. During formal triage with a dedicated nurse, results were similar to the original study.

However, on the afternoon that patients were triaged while being weighed (eyeballed), the percentage of patients admitted who were prioritised fell from 86% to 50%. An unpublished study (unpublished report 1997, Dr J Robson, consultant in paediatric A&E, V Lavy, general practitioner) carried out in the same department prior to triage training, showed that of 50 severely ill children requiring admission only 13 were recognised by nursing staff.

Little has been written about inter- and intraobserver variation in triage. One study in the USA,30 looking at five scenarios, showed inter-rater agreement to be poor (kappa value = 0.347 overall), and only 24% of participants rated the five cases the same severity in both phases. This lack of agreement in expected variations, and the lack of blinding of our assessors, makes it difficult to comment on the differences in category allocation in our study.

Correlation of triage category with likelihood of admission was good. Eighty three per cent of patients prioritised by the nurses during triage were admitted, and in the unsupervised triage this varied between 80% and 96% (excluding the afternoon of no formal triage). Admission rates for individual categories are comparable to those quoted for a recent implementation study of a new five level triage system in Boston.31

For 15 of the children who were triaged no record of their outcome was found. Tracing patients is a constant challenge because of the shortage of paper with overwriting on many records, and the lack of consistency in recording patients' details. Sometimes the child's date of birth is unknown, even to the mother, and many children pass through the department with the same, or similar names. We are, however, fairly confident of our mortality figures, as fatalities are presented each morning at a departmental meeting and these records are kept separately. The lack of such patient data also makes it impossible to know whether patients re-presented or died after discharge. Because of the absence of reliable prestudy data there is no evidence for improvement in mortality and morbidity in UFC during the study period.

As systems presently in use in the developed world have not been extensively validated there is no gold standard to compare with the performance of the ETAT guidelines. However, use of the guidelines appears to reliably select out patients needing admission. This applies to use by both nurses and APLS trained doctors, and is independent of concordance of priority signs.