Pertussis requiring intensive care

Methods

The audit included all PICU pertussis admissions from December 1991, when the unit opened, to December 2003. Hospital administrative approval and confirmation that ethical approval was not required were obtained.

Pertussis cases were identified from three sources. Firstly, a search of the PICU database, using the principal PICU discharge diagnosis of pertussis, identified patients. Second, using the same database, all discharge diagnoses of apnoea or pneumonia where pertussis had been queried were retrieved. These were included if there was:

• A positive culture or polymerase chain reaction (PCR) result for Bordetella pertussis. The PCR assay used primers selected from insertion sequence 481.¹²,¹³,¹⁴,¹⁵,¹⁶

• An illness consistent with pertussis for which no alternative cause was identified. This included apnoea and a paroxysmal cough with cyanosis or desaturation. Those with wheeze as a predominant symptom were excluded.

Third, the hospital laboratory provided a list of all B pertussis isolates (from 1996) and all positive B pertussis PCR assays (from 1998). These were cross‐checked with the PICU database.

Data were abstracted from the PICU database and the hospital records. Variables included those describing demographics, the presenting illness, pre‐hospital care, immunisations and household exposures, clinical status on presentation to the PICU, disease severity, treatment and complications.

The NZ immunisation schedule includes three infant pertussis vaccine doses, given at ages 6 weeks and 3 and 5 months. A 15‐month dose was added in 1996, acellular vaccine was introduced in 2000 and a 4‐year booster dose was included in 2002.¹⁷ Immunisation coverage surveys from the 1990s indicate that the three‐dose infant series was received by between 80% and 90% of children with 50%–60% receiving these immunisations on time.¹⁸,¹⁹

Weight for age z scores were calculated in Epi Info 2000 (CDC, Atlanta, GA) using the WHO/CDC reference data.²⁰ Peripheral leukocyte and lymphocyte counts were considered elevated if above the 95% confidence limits for age.²¹

An electronic database was created from the case record forms (Microsoft Access, Seattle, WA). Data analysis was performed using SAS version 9.1 (SAS Institute, Cary, NC). Time trends were examined using Poisson regression. Logistic regression was used to identify variables associated with death or disability. As the odds ratios were determined on a cohort in which death or disability was a relatively frequent outcome, the relative risk estimated from the odds ratio was adjusted to more accurately represent the true relative risk.²²

Results

There were 72 children. Pertussis was the PICU discharge diagnosis in 66. Another 38 in whom pertussis was queried were identified and five of these were included, four of whom were PCR positive. The PCR negative infant had paroxysms, whoop, seizures and a respiratory arrest. One other patient was identified from laboratory records (PCR positive).

The 33 of 38 children with PICU discharge diagnoses of apnoea or pneumonia in whom pertussis was queried were excluded as follows: PCR negative (11), culture negative but PCR not performed (three), wheeze (five), RSV identified but no laboratory investigations for pertussis (six), alternative cause for symptoms (two) and insufficient symptoms to diagnose pertussis (six).

Discussion

The atypical presentation of severe pertussis was apparent. Of the case‐defining symptoms, paroxysmal cough and apnoea were the most important diagnostically and in over half of the patients they were the only case‐defining symptom. Apnoea was frequently a presenting symptom in those with cough for 1 week or less. It is of note that the current Australian and American pertussis case definitions do not include apnoea.²⁹,³⁰ This study indicates that apnoea should be one of the case‐defining symptoms.

This retrospective study will have underestimated the true number of pertussis PICU admissions. Cases where pertussis was a differential diagnosis were excluded in some instances without adequate laboratory investigation. A recent prospective laboratory study from London identified pertussis in 20% of infants admitted to a PICU with respiratory failure, apnoea or bradycardia or an acute life threatening event.⁶

Death and disability occurred frequently. Many of the risk factors for these poor outcomes were already known.²,³¹ The increased risk of poor outcome associated with PICU readmission from the ward has not been described previously. PICU admission criteria were similar to those for any tertiary unit in the developed world and include recurrent apnoea or significant or insipient respiratory failure. Discharge from the PICU only occurred after a period of clinical stability. Thus, it seems unlikely that the need for readmission reflected local practice.

It is more likely that readmission occurred in those with more severe disease. That a smaller proportion of these patients had paroxysmal cough and that they initially presented with fewer case‐defining symptoms implies earlier presentation. The reasons for readmission were consistent with disease evolution. For national or regional PICUs, infants with possible pertussis should remain on the ward at the tertiary centre for longer than would otherwise be the case before returning to the referring hospital.

The annual number of pertussis PICU admissions increased from 1992 to 2003. PCR helped confirm cases in more recent years. However, the clinical illness was not different in those with or without laboratory confirmation of B pertussis infection. Hence PCR did not identify a milder spectrum of illness. This increase is consistent with the increasing pertussis hospitalisation rate in NZ over the past four decades, which was apparent well before PCR became available.⁷ Recent increases in severe disease in young infants have also been reported from the United States.⁴

This study had only modest power. The scope was also limited by the retrospective data collection and the incomplete follow‐up. This limited our ability to fully investigate ventilatory and circulatory support (eg, the contribution that mean airway pressure, and pulmonary and systemic arterial pressures make to outcome).⁴,³² Prospective enrolment into a multicentre study appears necessary.

How can we reduce severe pertussis in young infants? Immunisation reduces the risk of severe disease.¹⁰ Half of the children in this series who were old enough had received no immunisations. Thus the immediate intervention is on time delivery of the current immunisation schedule.

Other potential strategies include reducing the infectivity of contacts, by vaccinating family members and health care workers, and earlier commencement of immunisation. Acellular pertussis vaccines may enable immunisation schedules to start at, and protect from, a younger age.³³

Household contacts appeared to be the key source of infection for these infants. This wide age range of contacts is likely due to a combination of the mediocre immunisation coverage in New Zealand and the endemic nature of pertussis in adults, adolescents and school‐aged children.³⁴,³⁵,³⁶

The revision of the NZ immunisation schedule in 2006 to include a booster dose of pertussis vaccine at 11 years of age should reduce the infectivity of adolescents.³⁷ Subsequent boosters, either of the whole population or of expectant parents, should also be considered.

Infants die of pertussis despite ventilation, nitric oxide and inotropes. Only 30% of cases treated with ECMO survive.¹ Thus novel or disease specific treatments seem necessary. Recent reports indicate that haemofiltration, exchange transfusion and pulmonary vasodilators that do not act via the adrenergic system, namely milrinone and sildenafil, may be useful.³⁸,³⁹,⁴⁰ Risk factors for poor outcome identified on admission should guide subsequent PICU care and identify children for whom such interventions are most appropriate.

Supplementary Material