Rising National Prevalence of Life-Limiting Conditions in Children in England

Definition of Life-Limiting Conditions

An a priori list of ICD-10 codes that constituted the conditions of interest was produced before accessing the HES dataset. Two independent sources of information were used: the Hain Dictionary version 1.0 of ICD-10 codes for children seen by palliative care providers (developed by R.H., see Supplemental Information) and a list of diagnoses for children accepted for care at Martin House Children’s Hospice from 1987 to 2010. A 4-digit ICD-10 code was assigned to 92% of diagnoses on the Martin House list; the 8% not coded were children without clear diagnoses (eg, “degenerative neurologic disease with no firm diagnosis”).

Combining both sets of codes produced a provisional list of 801 ICD-10 codes for further scrutiny (84% of codes appeared on both lists).

All of these ICD-10 codes were individually subjected to the following 2 questions:

1. Are most children with this diagnosis life-limited/life-threatened?

2. Are most subdiagnoses within the ICD-10 code life-limiting/life-threatening?

A list of ICD-10 codes that fulfilled these criteria was compiled and completed by adding all malignant oncology ICD-10 codes (the data source was hospital admissions, so this would not include children “cured” of cancer).

The final ICD-10 coding framework consisted of 777 4-digit ICD-10 codes. Malignant oncology codes accounted for 445 (57%) codes, with congenital malformations and chromosomal abnormalities having 87 (11%) codes (available as Supplemental Table 3).

For statistical analysis, diagnoses were categorized into 11 groups based on the main ICD-10 chapters: neurology, hematology, oncology, metabolic, respiratory, circulatory, gastrointestinal, genitourinary, perinatal, congenital, and “other.” No attempt was made to prioritize multiple diagnoses for individuals; therefore, individuals may have more than 1 life-limiting diagnosis.

Patient Data

An extract of inpatient HES¹⁰ was obtained from the NHS information center for the 10-financial year time period 2000/2001 until 2009/2010. The selection captured all episodes for all patients ever coded, with 1 of the defined ICD-10 codes and/or the ICD-10 code for palliative care (used to capture children with no firm diagnosis). The extract excluded patients older than 19 years at the start of an episode and those whose country of residence was outside England.

The start age recorded at the first hospital episode in each year was used to assign the age category for each individual. Age was categorized into 5 groups: younger than 1 year, 1 to 5 years, 6 to 10 years, 11 to 15 years, and 16 to 19 years.

The data for each hospital episode included a code for gender and ethnicity. Gender was coded as male, female, or not known. Individuals with more than 1 recorded gender were assigned the most commonly recorded gender. Ethnicity within HES should be self-reported by the patient. Individuals with more than 1 ethnicity were assigned the most commonly reported ethnicity, unless the most common ethnicity was “not known.”¹¹ This ensured that the same code for an individual’s ethnicity was assigned to all episodes (ie, if coded white in 2001/2002 data and 2002/2003 but not known in 2003/2004, he or she would be counted as white in all years). The 16 census ethnic groups¹² were merged into 4 super-groups to avoid very small numbers in some groups: white (white: British; white: Irish; white: other white), South Asian (Asian or Asian British: Indian; Asian or Asian British: Pakistani; Asian or Asian British: Bangladeshi; Asian or Asian British: other Asian), black (black or black British: black Caribbean; black or black British: black African; black or black British: other black), Chinese and other ethnic groups (mixed: white and black Caribbean; mixed: white and black African; mixed: white and Asian; mixed: other mixed; Chinese and other ethnic group). More than 50% of the other ethnic group population in the 2001 census had been born in the Far East.¹³

Each individual was assigned a local authority district (LAD) and government office region (GOR) of residence based on their lower super-output area of residence. These assignments were done per year and if an individual moved a LAD within that year, the first LAD reported that year was used. This allowed the individual to be assigned a new LAD over the time period but not within a year.

An index of multiple deprivation 2007¹⁴ score was assigned to each individual based on their LAD. The index of multiple deprivation 2007 is an area-based score that combines many housing, social, and economic indicators to indicate the level of deprivation in each area. These scores were split into 5 equal categories based on the scores for the whole of England (20% of the local authorities in each category; category 1 is the most deprived and category 5 is the least deprived). Population data by ethnicity were unavailable for smaller geographical areas, so lower super-output area deprivation score could not be used in analyses investigating the joint effects of deprivation and ethnicity.

Population Data

Populations at risk were midyear estimates by age, gender, and ethnic group for local authorities in England obtained from http://ethpop.org/. This source has been used in preference to the subnational estimates produced by the Office of National Statistics, because the cohort component population estimation model¹⁵ incorporates more detailed demographic information by ethnic group in relation to newborns, mortality, and, most importantly, both subnational migration and international migration. Indeed, the Office of National Statistics has recently warned about the quality of their estimates.¹⁶

Analyses

Prevalence and 95% confidence intervals (CIs) per 10 000 population (aged 0–19 years) were calculated overall, for each year, for each ethnic group per year, for each geographical unit per year, and for the age groups per year and the diagnostic groups per year.

All data manipulation was undertaken in Microsoft SQL server 2008; statistical analyses were undertaken in Stata version 11 (Stata Corp, College Station, TX).