NHS Performance Indicators

Back to Previous Page	Contents : July 2000	Indicators Main Page
HA Level Indicators	Data Table	Specification

All indicators, where possible, use the 3 most recently available sets of data; e.g. HES based indicators use 1998-99, 1997-98 and 1996-97.

In the June 1999 High Level Performance Indicator (HLPI) and Clinical Indicator (CI) publications, direct standardisation was used to age standardise with the European standard population (ESP) as the standard. The indicators were not standardised for sex.

In this year’s Performance Indicators publication, there are three changes to the method used for standardisation:

Indirect standardisation is used instead of direct standardisation, and the ratios (and their confidence intervals) are then converted into absolute rates.
The indicators are standardised for age and sex, not just age.
For each indicator, the reference population used for standardisation will be the appropriate relevant population for that indicator, as opposed to the ESP. For the former HLPIs, this will be the actual HA populations. For the former CIs, hospital admissions (as appropriate for each indicator) for both the HAs and Trusts.

These changes bring the method of standardisation in line with that used for the Scottish and Welsh clinical indicators.

REASONS FOR CHANGE

The main reasons behind the change in standardisation methodology are:

Indirect standardisation is more robust with small numbers and avoids the distortions caused by direct standardisation based on unstable age-specific rates.
It is more flexible to future refinements, such as standardising for other factors such as deprivation or comorbidity.

As there are gender variations in health outcomes, person rates need to be standardised for age and sex.

The age distribution of a hospital population is different to that of a general population such as the ESP, hence the former is a more appropriate basis for standardisation for the CIs.

Note – The "Our Healthier Nation" indicators in Area 1 (1iv, 1v, 1vi, 1vii and 1viii) remain directly age- standardised to the European Standard Population in line with OHN policy. The use of the European Standard Population allows the possibility of European comparisons for these indicators.

METHOD

The age and sex specific rates for England are applied to the Health Authority/Trust age and sex specific populations, to generate the numbers of events (by age and sex) that would be expected in the Health Authority/Trust if it had the same rates as England.

The ratio of the total of these expected events to the total observed events for the Health Authority/Trust gives the indirectly standardised ratio for the Health Authority/Trust. The ratios are then converted into rates by multiplying them by the crude rate for England. The age groups used in the standardisation process are as follows:

<1, 1-4, 5-9, 10-14, …, 85+.

For the CIs, the age and sex specific rates for England for each indicator for each year are used as the reference for standardisation.

For example, for CI 1 the age and sex specific perioperative mortality rates for England in 1998-99 are applied to the age and sex specific numbers of inpatient spells with an operative procedure in each Health Authority/Trust in 1998-99 to obtain the number of "expected" deaths. For CI 4 the age and sex specific readmission rates for England are applied to the age and sex specific numbers of admissions in each Health Authority/Trust to obtain the number of "expected" readmissions. And so on for the other indicators.

NOTES:

Various age groups are used for the HLPIs and CIs (all ages, 0-15, 15-39, 35-74, 50+, 65+). The observed and expected values used to calculate the ratios (see the worked example below) relate to the age group as appropriate for each indicator.

In the case of the HLPIs, the standard population from which average occurrence rates are calculated will be the actual England population. To obtain a fixed base to enable comparisons overtime, ONS mid-year resident population estimates for 1998 by HAs (April 1999 boundaries) along with England occurrences for that indicator for 1998/99 will be used to calculate the expected number of events for each HA during the standardisation of all 3 years of data. In practice, this means using mid 1998 populations and 1998/99 occurrences in the standardisation of 1996/97, 1997/98 and 1998/99 data.

Unlike the HLPIs that measure population-based risk, the CIs (Health Authorities and Trusts) are hospital population based and not general population based. For CIs the numerators and denominators are therefore the patient-specific populations as appropriate for each indicator (see para 4). The numerators and denominators for England, Health Authorities and Trusts used in the standardisation process will therefore vary from indicator to indicator.

Unlike HLPIs, CIs are standardised within an individual year (1996/97, 1997/98, 1998/99). This means that the standardised cannot be compared across time, but can be compared within an individual year. The calculations were carried out in this way due to concerns about the data quality of earlier years.

The CIs are as follows:

CI 1: Rates of deaths in hospital within 30 days of surgery (all ages) by method of admission

CI 2: Rates of deaths in hospital within 30 days of emergency admission with a hip fracture (neck of femur), for patients aged 65 and over

CI 3: Rates of deaths in hospital within 30 days of emergency admission with a heart attack (myocardial infarction), for patients aged 35-74

CI 4: Rates of emergency readmission to hospital within 28 days of discharge from hospital (all ages)

CI 5: Rates of discharge to usual place of residence within 56 days of emergency admission from there with a stroke, for patients aged 50 and over

CI 6: Rates of discharge to usual place of residence within 28 days of emergency admission from there with a hip fracture (neck of femur), for patients aged 65 and over.

Worked examples are given below for an HLPI and a CI.

HLPIs: Worked Example

Consider the following data for HLPI 3vi, component 1 for HA X:

Age range	HA asthma admissions, x(j)		HA population, n(j)		England asthma admissions, xs(j)		England population, ns(j)
j	Male	Female	Male	Female	Male	Female	Male	Female
0	15	7	2453	2014	1790	659	306982	292343
1-4	89	45	9344	9112	11167	5643	1269671	1205375
5-9	35	29	11951	12045	4294	2376	1672504	1591836
10-14	20	19	9971	9765	2682	2177	1612712	1528465
15-19	4	17	10276	9943	1180	2246	1558288	1474536
20-24	2	11	11080	11201	1045	2213	1494077	1427002
25-29	4	9	14589	13943	1256	2300	1890723	1794013
30-34	12	11	16778	15643	1200	2188	2087414	1992725
35-39	10	29	15032	14157	931	1971	1948348	1878082
40-44	6	11	12834	11693	855	1714	1655145	1635575
45-49	3	8	10907	11219	872	1728	1606494	1603542
50-54	8	12	10210	10845	907	1545	1647615	1653993
55-59	8	7	6995	9176	832	1455	1265270	1282857
60-64	4	9	7234	7634	818	1264	1150518	1189447
65-69	2	11	5978	7126	782	1331	1032916	1140345
70-74	2	15	4003	5432	657	1116	880282	1073878
75-79	3	5	3793	5783	458	959	683480	987552
80-84	8	2	1632	3829	299	673	361739	661470
85+	2	4	1732	3561	133	492	253860	703508
Total	237	261	166792	174121	32158	34050	24378038	25116544

Expected asthma admissions are then calculated for each sex and age group using the formula:

Expected no. admissions =

where n(j) = population in age group j in HA X

xs(j) = number of occurrences in age group j in the standard population

ns(j) = population in age group j in the standard population

So for age 0, expected no. male admissions = 2453*(1790/306982)

= 14.3

expected no. female admissions = 2014*(659/292343)

= 4.5

The full set of expected admissions is shown in the table overleaf:

Age range	Expected admissions e(j)
j	Male	Female
0	14.3	4.5
1-4	82.2	42.7
5-9	30.7	18.0
10-14	16.6	13.9
15-19	7.8	15.1
20-24	7.7	17.4
25-29	9.7	17.9
30-34	9.6	17.2
35-39	7.2	14.9
40-44	6.6	12.3
45-49	5.9	12.1
50-54	5.6	10.1
55-59	4.6	10.4
60-64	5.1	8.1
65-69	4.5	8.3
70-74	3.0	5.6
75-79	2.5	5.6
80-84	1.3	3.9
85+	0.9	2.5
Total	226.0	240.5

The standardised ratio (SR) for persons is then calculated as:

SR = =

Where x(j) = number of observed occurrences in age group j for HA X

e(j) = number of expected occurrences in age group j for HA X

So in the example, SR = (237+261)/(226.0+240.5)

= 1.068

The final indicator value is then produced by multiplying through by the England crude rate per 100,000:

indicator value = SR * England crude rate

So in the example, England crude rate = ((32158+34050)/(24378038+25116544))*100000

= 133.8

Hence indicator value per 100,000 = 1.068 * 133.8

= 142.9

This crude rate multiplication is used in both the Scottish and Welsh clinical indicators and is done in order to produce a more useable and interpretable final value.

(Note – For the HLPIs for all 3 years worth of data (1996/97, 1997/98 and 1998/99) the standardised ratio will be multiplied through by the crude rate for England for 1998/99.)

CIs: Worked Example

Consider the following data for CI 2 (rates of deaths in hospital within 30 days of emergency admission with hip fracture) for patients aged 65 and over:

Age range	HA/Trust X				England
	Numerator		Denominator		Numerator		Denominator
	Denom spells where patient dies <30 days x(j)		Spells with emer admission for hip fracture n(j)		Denom spells where patient dies <30 days xs(j)		Spells with emer admission for hip fracture ns(j)
j	Male	Female	Male	Female	Male	Female	Male	Female
0	0	0	0	0	0	0	5	8
1-4	0	0	0	0	0	1	33	32
5-9	0	0	0	0	0	0	30	23
10-14	0	0	0	1	0	0	85	38
15-19	0	0	1	1	0	0	66	8
20-24	0	0	0	3	0	0	69	13
25-29	0	0	1	1	0	1	82	29
30-34	0	0	0	0	0	1	97	29
35-39	0	0	3	0	2	0	133	46
40-44	0	0	2	1	1	1	138	70
45-49	0	0	2	0	1	0	190	115
50-54	0	0	1	2	7	4	267	313
55-59	0	0	4	4	11	5	354	425
60-64	0	0	6	14	8	13	439	813
65-69	0	1	9	33	24	45	695	1657
70-74	0	1	10	35	66	133	1098	3315
75-79	0	5	20	76	197	330	1778	6514
80-84	6	10	22	93	270	531	1877	8451
85+	8	23	34	172	635	1861	2949	16782
Total	14	40	115	436	1222	2926	10385	38681
Total 65+	14	40	95	409	1192	2900	8397	36719

Expected hip fracture admissions are calculated for each sex and age group using the formula:

Expected no. admissions =

n(j) = population in age group j in HA/Trust X

xs(j) = number of occurrences in age group j in the standard population

ns(j) = population in age group j in the standard population

So for age 80-84, expected no. male admissions = 22*(270/1877)

= 3.16

expected no. female admissions = 93*(531/8451)

= 5.84

The full set of expected admissions is shown in the table overleaf:

Age range	HA/Trust X
Age range	Expected admissions e(j)
j	Male	Female
0	0.00	0.00
1-4	0.00	0.00
5-9	0.00	0.00
10-14	0.00	0.00
15-19	0.00	0.00
20-24	0.00	0.00
25-29	0.00	0.03
30-34	0.00	0.00
35-39	0.05	0.00
40-44	0.01	0.01
45-49	0.01	0.00
50-54	0.03	0.03
55-59	0.12	0.05
60-64	0.11	0.22
65-69	0.31	0.90
70-74	0.60	1.40
75-79	2.22	3.85
80-84	3.16	5.84
85+	7.32	19.07
Total	13.94	31.41
Total 65+	13.61	31.07

The standardised ratio (SR) for persons aged 65 and over is then calculated as:

SR = =

Where x(₆₅₊) = number of observed occurrences in age group 65+ for HA/Trust X

e(₆₅₊) = number of expected occurrences in age group 65+ for HA/Trust X

So in the example, SR = (14+40)/(13.61+31.07)

= 1.2086

The final indicator value is then produced by multiplying through by the England crude rate per 100,000:

indicator value = SR * England crude rate

So in the example, England crude rate = ((1192+2900)/(8397+36719))*100000

= 9069.95

Hence indicator value per 100,000 = 1.2086*9069.95

= 10961.9

This crude rate multiplication is used in both the Scottish and Welsh clinical indicators and is done in order to produce a more useable and interpretable final value.

Confidence Intervals

The lower limits (LL) and upper limits (UL) for the SMR are obtained prior to multiplying through by the crude rate. By exploiting the general relationship between confidence limits and test statistics, the limits may be found by solution of equations involving Poisson probabilities, for which Byar’s approximation is sufficiently accurate. For a 95% confidence interval, Byar’s approximation gives:

LL = * 100 *(1 - - )^3

UL = * 100 * (1 - + )^3

where x is the observed number of events and e is the expected number of events.

NHS Performance Indicators : July 2000ANNEX A Standardisation Methodology

NHS Performance Indicators : July 2000
ANNEX A Standardisation Methodology