Changes in Profile of Paediatric Intensive Care Associated With Centralisation

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Intensive Care Med (2001) 27: 1670±1673

DOI 10.1007/s001340101072 B R I E F RE PO RT

G. Pearson Changes in the profile of paediatric


P. Barry
C. Timmins intensive care associated with
J. Stickley
M. Hocking centralisation

Received: 16 January 2001


Abstract Objectives: To compare The proportion of admissions to the
Final revision received: 25 May 2001 intensive care admissions from a de- principal paediatric intensive care
Accepted: 30 July 2001 fined population of children in 1991 unit increased from 60 % to 90 %
Published online: 23 August 2001 and 1999, during a period of organi- (p < 0.0001) in association with its
 Springer-Verlag 2001 sational change and centralisation of expansion from 6 to 18 beds. Length
paediatric intensive care. of ICU stay decreased from 103 to
Funded in part by West Midlands Regional
Design: Two 12-month population- 74 h (difference 29 h, 95 %CI,
Health Authority based audits were compared. Data 4.78±54.2 h, p = 0.0117). Child mor-
were collected from hospitals in Bir- tality fell over this period by 34
mingham and the surrounding dis- deaths per 100,000 children (95 %CI
tricts. Denominator data were ob- 16±51, p < 0.0001). The proportion
tained from the Office for National of children requiring mechanical
Statistics. The place and rate of in- ventilation at admission to intensive
tensive care admission, the use of care was unchanged.
mechanical ventilation at admission, Conclusions: Centralisation by ex-
mortality and length of stay were pansion of the lead centre was asso-
compared. ciated with a large increase in the
Setting: Hospitals in the West Mid- numbers of children receiving in-
lands. tensive care consistent with an un-
Participants: All children (< 15 yrs) met need for paediatric intensive
living in Birmingham who received care in 1991, which may still exist.
)
G. Pearson ( ) ´ P. Barry ´ C. Timmins ´
J. Stickley ´ M. Hocking
intensive care during the study peri-
ods.
Centralisation of paediatric inten-
sive care may have contributed to
Birmingham Children's Hospital, Measurements and results: The the fall in child mortality over this
Steelhouse Lane, Birmingham B4 6NH, number of Birmingham resident time period.
UK
E-mail: Gale.pearson@bhamchildrens.
children admitted for intensive care
wmids.nhs.uk increased from 277 to 510 Keywords Intensive care
Phone: +44-1 21-3 33 96 71 (p < 0.0001) i.e. from 1.3 to 2.3 ad- utilisation ´ Bed occupancy ´ Health
Fax: +44-1 21-3 33 96 51 missions per 1,000 children per year. services research ´ England

a difference in the incidence of critical illness in children


Introduction
in the United Kingdom or a failure of provision in terms
In 1993 it was asserted that paediatric intensive care ser- of numbers or organisation of intensive care beds [5]. At
vices in the United Kingdom were inadequate [1, 2, 3]. that time relatively few children received intensive care
Admission rates to intensive care were very low com- in dedicated paediatric intensive care units, and a large
pared to those in Australia and the United States [4]. number of general intensive care units provided inci-
There was considerable debate as to whether this re- dental intensive care for children as and when required.
flected a higher threshold for intensive care admission, In 1996 a `Framework for the Future' [6] was introduced
1671

Table 1 Children's use of intensive care in Birmingham: 1999 compared to 1991


1991 1999 Absolute Ratio p
change
Children resident in Birmingham 212,700 222,200 9,500 1.04
Number of deaths 221 156 ±65 0.71 < 0.0001
Birmingham child residents enter study 297 590 293 1.99 < 0.0001
Birmingham child residents get admitted to an ICU 277 510 233 1.84 < 0.0001
Birmingham child residents > 28 days old admitted to ICU 245 427 182 1.74 < 0.0001
Child ICU admissions per 1,000 children in the population 1.30 2.30 1 1.77
Number of deaths in the study 28 34 6 1.21 0.051
Number of deaths occurring in ICU 26 32 6 1.23 0.080
Number of deaths in ICU (excluding neonates, £28 days at admission) 24 26 2
Number of sites where children admitted to intensive care 10 10 0 1
ICU mortality (%) 9.39 6.27 ±3 0.67
Non-neonatal ICU mortality 9.8 6.0 3.8 0.61
Percentage of all deaths occurring in ICU 11.76 20.51 9 1.74
Percentage of ICU admissions occurring to a PICU 61 90.20 29.2 1.58 < 0.0001
ICU child deaths per 100,000 children in the population 12.22 14.40 2.18 1.18
Deaths per 1000 children in the population 1.04 0.70 ±0.34 0.67
Percentage male 57 57 ± ±
Median age (years) 1.58 1.26 ± ± 0.787

that planned increasing centralisation of paediatric in- for severity of illness to detect a change in the pattern of care from
tensive care in `Lead centres'. By 1999 some of these `high dependency care' to `intensive care'. This assumes that the
threshold for initiating mechanical ventilation remains the same.
changes in configuration had been implemented in Bir-
The significance of differences between the two time periods were
mingham. We therefore analysed population-based calculated using the c2 statistic (with Fisher's exact test when fre-
data from 1991 and 1999 to assess any associated change quencies were small). Confidence intervals for differences in pro-
in paediatric intensive care activity and outcome. portions were calculated using confidence interval analysis soft-
ware version 1 [8].

Method
Patients entered the 1991 study [7] if they were aged less than
Results
15 years on admission to hospital and received intensive care. The Between 1991 and 1999 the number of beds available in
data were collected throughout the study period by the principal
researcher (a physician in training) visiting hospitals in the four
the principal paediatric intensive care unit increased
Birmingham Health Authorities and each of the neighbouring from 6 to 18. Additionally a second paediatric intensive
health authorities. The data for the year 1999 was extracted from care unit opened in a district general hospital north of
a larger database collected prospectively from all hospitals in the the city. There was an associated increase in use of in-
West Midlands region by ward nursing staff coordinated by a full- tensive care by Birmingham children from 277 admis-
time researcher. The entry criteria used in the 1991 study were ap- sions to 510, or from 1.3 to 2.3 admissions per thousand
plied to the 1999 database to give comparative information. Data
from hospitals not included in the former study were excluded. `In-
children and 29 % more children received their inten-
tensive care' was defined as admission to an intensive care unit or sive care in a paediatric intensive care unit (PICU;
the use of a `unique intensive care treatment' (mechanical ventila- 95 %CI 23±35 %, p < 0.0001). The change in activity be-
tion, continuous vascular or intracranial pressure monitoring or va- tween the two epochs is shown in Table 1, with the level
soactive drug infusion). Children on neonatal intensive care units of significance where appropriate. Table 2 shows the
who had not been discharged from hospital since birth were ex- distribution of the deaths between adult and paediatric
cluded. Residence within Birmingham was subsequently deter-
mined by the patient's postcode. Denominator data for both ep-
intensive care units in the two epochs. The predominant
ochs were provided by the Office for National Statistics. increase in activity occurred in the lead centre, which
The incidence, pattern of provision and outcome of critical care delivered 61 % (169 patients) of Birmingham children's
for children resident in Birmingham were compared. The degree intensive care in 1991 compared to 90 % (460 patients)
of centralisation was expressed as the proportion of the total num- in 1999. Only 9 Birmingham residents were admitted to
ber of intensive care admissions occurring in the largest (highest the new unit outside the city in the second study period.
volume of child admissions) intensive care unit. The principal pae-
diatric intensive care unit was known to be the same in the two ep-
Furthermore, two hospitals stopped delivering intensive
ochs. It serves a larger population than the residents of Birming- care to children in the interval between the studies, one
ham. The use of mechanical ventilation at admission to intensive of which closed altogether. Its workload transferred to
care was known for both time periods. This was used as a surrogate other hospitals remaining in the study. Of the absolute
1672

Table 2 Distribution of patients between adult and paediatric in- Table 3 Projected intensive care bed requirements for England
tensive care units: Birmingham residents and Wales
1991 1999 Model 1 Model 2 Model 3
Admis- Deaths Admis- Deaths Admission rate per thousand 2.3 1.55 1.55
sions sions children
Birmingham Chil- 169 14 451 32 Mean length of stay (Days) 3.08 3.08 2.1
dren's Hospital PICU Number of PICUs 16 16 16
Other PICU 0 0 9 0 Mean occupancy 0.8 0.7 0.8
Adult intensive 108 12 50 0 Beds per PICU 22 18 16
care unit

increase in activity for Birmingham residents shown in direction with those of a previous study in which centra-
Table 1, 70 % was due to new `general paediatric' admis- lised and decentralised paediatric intensive care were
sions. The greatest absolute increases in admission fre- compared [9]. The excess deaths in the decentralised
quency were those for respiratory and neurological di- system in that study accounted for a large proportion of
agnoses. As might be anticipated, the number of cardiac the difference in child mortality rates between the par-
and neonatal surgical cases were relatively fixed. ent populations. In this study, larger differences in child
The use of mechanical ventilation on admission to in- mortality occurred (33 % reduction as opposed to
tensive care was the same in the two periods (78 % in 11 %), but other factors will have contributed to the
1991 and 81 % in 1999, p = 0.36). As there was no major change. This study also included neonates (excluded in
change in policy regarding the use of mechanical venti- the former study), but the differences in ICU mortality
lation, this finding supports the premise that there was persist regardless of whether neonates are included (Ta-
no large change in severity of illness between the two ble 1).
periods. However, the overall child mortality rate fell As in the former study, this comparison used popula-
by 0.34 deaths per 1000 children (95 %CI 0.16±0.51, tion-based data in preference to unit-based data. This
p < 0.0001) which is 75 less child deaths (based on the approach is preferred because the results are less sus-
1999 population size). Patients stayed in intensive care ceptible to case-mix effects such as those related to the
for a shorter period of time. The mean length of stay referral status of the hospitals involved. The apparent
fell from 103 h to 74 h, a difference of 29.5 h (95 %CI beneficial effects of centralisation could still be due to
4.78±54.2 h, p = 0.0117). a positive volume:outcome relation [10] both within the
lead PICU alone, which actually admitted a total of
1278 patients during the 1999 period of data collection
versus 713 in 1991, and in the moving of children from
Discussion
small volume units to the large volume unit. Fifteen of
This analysis of population based data from 1991 and the fifty children admitted to adult intensive care units
1999 demonstrates a virtual doubling of intensive care in 1999 were subsequently transferred to the lead cen-
use by Birmingham children. Childhood mortality de- tre. Furthermore, the improved outcome may be occur-
creased in the same time period. Improved performance ring despite the influence of the hospitals training status
of intensive care for children is implied by the lower in- [11].
tensive care unit mortality and the shorter length of Over the period 1991±1999 the national child mortal-
stay, but the former change is not quite significant be- ity rate (England and Wales) decreased by the same
cause of the low number of deaths. amount as that for Birmingham residents (ratio of stan-
The number of children resident in Birmingham who dardised ratios 1.01, 95 % confidence interval
were admitted to intensive care increased significantly 0.852±1.18). Changes in Birmingham might therefore
between the two epochs and was higher in 1999 than in mirror those in other regions. However, the degree of
the West Midlands region as a whole (which was 1.55 centralisation evident in Birmingham in 1991 was al-
per thousand children). This could reflect a change in ready much greater than that in other cities [3, 9] and re-
the severity of illness of the children or a change in the gional variations may persist.
threshold for intensive care admission. It is more likely Data collected in 1991 did not include a risk adjust-
to reflect partial resolution of an under-provision of ment tool, and the tool used in the 1999 study [12] was
paediatric intensive care. It is not possible to distinguish not available in 1991. Risk adjustment techniques, how-
the major factor directly. ever, assume that intensive care is provided and to the
These results do not prove a causal link between cen- standard that applied in their derivation datasets. They
tralisation and reduced mortality, but they do concur in do not help interpret the outcome for children who do
1673

not receive intensive care, which was the accusation in projected mean occupancy rates were used. Model 3,
1993 [1, 2]. Nevertheless, even though the strength of for interest, uses the length of stay from Victoria, Aus-
this comparison is arguably reduced by the lack of ªse- tralia, in 1994 [9].
verity of illnessº data for the 1991 epoch, the need for
ventilation serves as a surrogate for severity of illness.
The fact that the incidence of mechanical ventilation at
Conclusion
the outset of intensive care did not change, argues
against a change in the threshold for intensive care ad- Over the period 1991±1999 there was an increased cen-
mission. There was either a lack of provision of inten- tralisation of paediatric intensive care for children living
sive care for children in 1991 or by 1999, a change in ad- in Birmingham. This was achieved by increasing the
mission threshold for intensive care had occurred ac- numbers of beds available in the lead centre. The num-
companied by an increase in this intervention. ber of children receiving intensive care rose, and child
The number of PICU beds required to serve the pop- mortality and length of intensive care stay fell over the
ulation of England and Wales (10 million children) can same period. These data are consistent with an unmet
be extrapolated from the 1999 data by approximation need for paediatric intensive care in 1991, which may
to a Poisson distribution [6, 13]. Table 3 shows three still exist. Centralisation of paediatric intensive care
such models. Models 1 and 2 were derived using the may have contributed to the fall in child mortality over
length of stay and admission rates reported here for Bir- this time period.
mingham and the West Midlands as a whole. Typical

References
1. Shann F (1993) Paediatric intensive 6. National Coordinating Group on Paedi- 10. Tilford JM, Simpson PM, Green JW,
care. Lancet 342: 1240 atric Intensive Care (1997) Paediatric Lensing S, Fiser D (2000) Volume-out-
2. Shann F (1993 Australian view of paedi- intensive care ªA framework for the fu- come relationships in pediatric inten-
atric intensive care in Britain. Lancet tureº. NHS Executive, London sive care units. Pediatrics 106: 289±294
342: 68 7. Barry PW, Hocking MD (1994) Paedi- 11. Pollack MM, Patel KM, Ruttimann E
3. British Paediatric Association (1993) atric use of intensive care. Arch Dis (1997) Pediatric critical care training
The care of critically ill children: report Child 70: 391±394 programs have a positive effect on pedi-
of a multidisciplinary working party on 8. Gardner SB, Winter PD, Gardner MJ atric intensive care mortality. Crit Care
intensive care. British Paediatric Asso- (1989) Statistics with Confidence. In: Med 25: 1637±1642
ciation, London Gardner MJ, Altman DG (eds) Statis- 12. Shann F, Pearson G, Slater A, Wilkin-
4. British Paediatric Association (1987) tics with confidence. BMJ, London son K (1997) Paediatric index of mor-
Report of a working party on paediatric 9. Pearson G, Shann F, Barry P, Vyas J, tality (PIM): a mortality prediction
intensive care. British Paediatric Asso- Thomas D, Powell C, et al (1997) model for children in intensive care. In-
ciation, London Should paediatric intensive care be cen- tensive Care Med 23: 201±207
5. Sheldon T (1995) Which way forward tralised? Trent versus Victoria. Lancet 13. Milne E, Whitty P (1995) Calculation of
for the care of critically ill children? 349: 1213±1217 the need for paediatric intensive care
University of York, York beds. Arch Dis Child 73: 505±507

You might also like