Results

Over the 12 months of the study, there were 35 918 attendances to the children's accident and emergency department, of which 9239 were for a medical condition. A total of 233 (2.5%) children who presented to the department had a petechial or purpuric rash. We excluded 15 children who had a clear alternative diagnosis (11 with Henoch–Schonlein purpura, one with idiopathic thrombocytopenic purpura, one with haemolytic uraemic syndrome, one with acute leukaemia, and one with a previously recognised clotting disorder), leaving a total of 218 study children. Twenty four of the 218 children (11%) had proven meningococcal disease. Table 1 shows the serogroups of the meningococci and mode of diagnosis. A further four children had possible meningococcal disease with a non-blanching rash. Two had raised antibody titres to meningococcal outer membrane proteins with a greater than fourfold rise in convalescent titres, and one had a positive throat swab. A fourth child with a widespread purpuric rash required ventilation and inotropic support. She had received intramuscular benzylpenicillin before arrival and her blood culture, PCR, serology, and throat swabs were negative. Since the diagnosis was unproven, these children were all included in the non-meningococcal group for analysis. No child had laboratory confirmation of bacteraemia with any other bacteria. Eight children (3.7%) did not have blood cultures taken—they were not were treated with antibiotics and did not develop signs of sepsis. Six children were admitted with proven invasive meningococcal infection (five with meningitis) in the same 12 month period but did not have a non-blanching rash.

Neither season nor age was useful in predicting meningococcal infection. Fifty five per cent of children with a non-blanching rash were less than 3 years old (median age 2 years) and meningococcal infection was also more common in younger children (median age 2 years). More children with a non-blanching rash were seen in the winter months (December to February) than in the other seasons (χ²; p < 0.001); although meningococcal disease was more common in the winter months, this was not statistically significant (χ²; p = 0.3). A total of 184 children (84%), including all 24 who were later proven to have meningococcal infection, were admitted to hospital for a median time of 24 hours. One child who was clinically well was admitted to hospital but discharged with no treatment—blood culture grew N meningitidis at 48 hours. She was well when she was recalled and repeat blood cultures prior to initiation of treatment were negative. A total of 101 children (46%) received antibiotics (96% intravenous, 4% oral). No child was sent home from the accident and emergency department and subsequently readmitted with meningococcal infection. One child died from meningococcal infection during the study period.

Children with meningococcal infection were more likely to be ill (OR 16.7; 95% CI 5.8 to 47.6), to have an axillary temperature >38.5°C (OR 8.0; 95% CI 2.7 to 23.8), purpura (OR 37.2; 95% CI 11.7 to 118.3), and a capillary refill time of more than two seconds (OR 29.4; 95% CI 9.4 to 92.6) than non-meningococcal children, although a substantial minority of children without meningococcal disease showed these features (table 2). Hypotension was more common in those with meningococcal disease but blood pressure was only measured in a third of all children. It is likely that these were selectively more unwell. No child with a rash confined to the distribution of the superior vena cava (head, neck, and chest above the nipple line) (74/218) had meningococcal infection.

Children with meningococcal infection were more likely to have an abnormal neutrophil count (OR 2.7; 95% CI 1.1 to 6.5) and a prolonged INR (OR 30.0; 95% CI 9.9 to 91.0). However, a substantial minority of children without meningococcal disease also showed these features (table 3). No child with a CRP of less than 6 mg/l (90/183) had meningococcal infection.

Several of the clinical features and investigations were sensitive but poorly specific, while others were more specific but insensitive as predictors of meningococcal infection (tables 4 and 5). A rash confined to the distribution of the superior vena cava and a normal CRP each had a negative predictive value of 100% but no feature had a high positive predictive value. If the data are reanalysed, classing the four suspected cases described above as having meningococcal rather than non-meningococcal disease, the 100% negative predictive value and sensitivity of a rash in the superior vena cava distribution and a normal CRP are unchanged. Purpura, delayed capillary refill, hypotension, abnormal INR, and an abnormal neutrophil count become more specific but no less sensitive as predictors of meningococcal disease.

Discussion

We have shown that 11% of all children who presented with a petechial rash had meningococcal infection. Our incidence of meningococcal infection is high, reflecting the known recent increase in the UK.4 The study was carried out just before the national immunisation campaign against meningococcal C infection was launched. Assuming that the campaign is effective, our study predicts that this incidence figure will halve. This has implications for the predictive values as the negative predictive value of our observations will increase and the positive predictive value will decrease if the incidence of meningococcal disease falls. Eighty nine per cent of children did not have meningococcal infection. In 34% of children who presented with a petechial rash, the lesions were confined to the distribution of the superior vena cava. The rash is probably caused by raised venous and capillary pressure resulting from coughing, vomiting, or crying. The finding has been noted by others.5 ,6 Most of these children were well and none had meningococcal infection. In the remaining 55% of children with a petechial rash beyond the SVC distribution but without meningococcal infection, the rash was presumably caused by vasculitis or disordered coagulation in association with other non-bacteraemic bacterial or viral infections.

Other investigators have examined the incidence of meningococcal disease in children with fever and petechiae but not petechiae alone. Nguyen et al studied 129 children with fever and petechiae in New York in 1984 by retrospective case note review.7 Twenty per cent had culture proven bacterial infections, 11% with N meningitidis. No single laboratory investigation was sufficiently sensitive to detect all patients with life threatening bacterial infections, but a combination of normal laboratory findings was highly specific for those without significant disease. Baker et alstudied 190 children with fever and petechiae prospectively over 12 months in Cincinnati in 1989.8 Seven per cent had meningococcal disease. They concluded that patients with petechiae on their lower extremities were more likely to have invasive disease. Those with petechiae above the nipple line only were unlikely to have invasive disease. No variable was sensitive enough to detect all patients with serious disease. Mandl et alconducted a prospective study of 411 children presenting with fever and petechiae to a large emergency department in Boston over 18 months.9 Only five had proven sepsis, two withN meningitidis, a very low incidence. They concluded that in their population, children with fever and petechiae who appeared well could be treated as outpatients because the few children with bacterial sepsis were all ill. Kuppermannet al compared clinical and haematological features of children with undiagnosed meningococcal disease to other febrile children.10 They found no clinical or haematological feature to be routinely helpful in the diagnosis of unrecognised meningococcal disease.

Brogan and Raffles recently reported the findings in 55 children presenting with fever (>37.4°C) and a petechial rash, three of whom had meningococcal disease.11 They proposed the “ILL” criteria (irritability, lethargy, and low capillary refill time) in association with an abnormal total white cell count and abnormal CRP. Our study supports their suggestion that illness and delayed capillary refill time are predictors of meningococcal disease, but not their findings with regard to an abnormal total white cell count. Brogan and Raffles included only children presenting with a temperature >37.4°C in association with a petechial rash, suggesting that fever is an essential component of the illness. Other studies have reported on the combination of fever and rash too.7-10 Our study is the first to be based on the presence of a non-blanching rash alone, regardless of fever. Five of our children with meningococcal disease (21%) did not have an axillary temperature of 37.5°C or more. Clearly they may have had fever before or after admission, but lack of fever at the time of assessment does not exclude meningococcal disease.

There are possible limitations to our study. We believe we identified all children with petechiae by cross checking the completed proformas with all investigations sent for a petechial rash and all admissions for a petechial rash. It is possible, though, that a child who was completely well with a petechial rash may have been sent straight home without being identified. It is very unlikely that any cases of meningococcal infection were missed, as all the blood cultures from the Nottingham area are processed through the same laboratory and our cases were cross checked with these. It is possible, though unlikely, that a child who was discharged, later presented further afield. Our children's intensive care unit is the referral centre for the surrounding area and any ill child with meningococcal sepsis would have been referred there. Although a high percentage of children were investigated, data were incomplete, limiting some of the conclusions. Data collection was by different grades of doctor (junior house officer 10%, senior house officer 42%, and specialist registrar 48%)—ideally this would have been carried out by a single investigator. However, it means that the study reflects actual practice rather than an ideal.

We have shown that clinical features are useful predictors of meningococcal disease. Of 218 children with a petechial rash, only one (3%) who was well had meningococcal disease compared with 35% who were ill at presentation. Forty seven per cent of children with purpura (>2 mm) had meningococcal disease compared with 2% of children with petechiae alone. Forty two per cent of children with a prolonged capillary refill time had meningococcal disease compared with only 2% of those with a normal time. Low blood pressure was a late sign—only five children with meningococcal disease were hypotensive. As discussed above, fever was a less helpful indicator as 21% of those with meningococcal infection were afebrile (<37.5°C), although meningococcal disease was more likely in those with an axillary temperature >38.5°C.

Our findings suggested that a child presenting with petechiae in the superior vena cava distribution is unlikely to have meningococcal disease. No child who presented with petechiae in the superior vena cava distribution alone had meningococcal disease—there is a 95% chance that only 0–5% of children with such a rash will have meningococcal disease. These children accounted for 34% of the total. Those with purpura, ill appearance, increased capillary refill time, or hypotension should be admitted and treated for meningococcal infection without delay. This accounts for a further 36%. By our estimates, 70% of these will not have meningococcal disease but 96% of meningococcal infection lies within this group. The other 4% of children with meningococcal infection lay within the remaining 30% of all the children. They were well, with or without fever, with a petechial rash which extended beyond the territory of the superior vena cava. Clinical features alone provide no guidance on their identification or management.

Investigations were less useful. Total white count did not distinguish between those with and without meningococcal disease. Abnormal neutrophil count and abnormal platelet count were slightly more useful (OR 2.7; 95% CI 1.1 to 6.5 and 3.9; 1.3 to 11.5 respectively). Clotting studies were more specific but lacked sensitivity. An abnormal INR was found in 58% of patients with meningococcal disease but only 5% of those without (OR 30.0; 95% CI 9.9 to 91.0). A normal CRP appears to be a useful predictor of those who do not have meningococcal infection. No child with a CRP < 6 mg/l (our laboratory definition of normal) had meningococcal disease—there is therefore a 95% chance that only 0–4% of children with a normal CRP will have meningococcal disease. We recognise the possible limitations of a normal CRP, especially in early disease. However, Marzouk et al, in a large study of 124 children with meningococcal disease, reported the diagnostic value of CRP measurement.12 Only eight of 124 patients had a CRP < 20 mg/l and of these, five patients had symptoms of less than 12 hours duration. They do not report how many of these had a CRP of < 6 mg/l. If CRP measurement is rapidly available at all times, our study suggests that it may provide a good guide to the identification of the minority of children with meningococcal disease who are well.

Our admission rate for children presenting with a petechial or purpuric rash was 84%. If we had discharged all those with a petechial rash confined to the superior vena cava distribution we would have reduced this rate to 68% (as 44% of those already discharged had a petechial rash in the superior vena cava distribution alone). If we had then discharged those remaining children who were well, with petechiae beyond the superior vena cava distribution, who did not have delayed capillary refill time or hypotension (that is, none of the features strongly associated with meningococcal disease) and who had a completely normal CRP, we would have reduced the admission rate further to 57% without sending home any child with invasive meningococcal disease.