3.2 Emergency readmissions within 30 days of discharge from hospital - Specification v1.4

Technical description

Indirectly age, sex, method of admission and diagnosis/procedure standardised percentage of emergency admissions to any NHS or independent sector hospital undertaking NHS commissioned activity in England occurring within 30 days of the last, previous discharge from hospital after admission.  It includes activity in English NHS Hospitals and English NHS commissioned activity in the independent sector

Admissions for cancer and obstetrics are excluded.

(Note: when using indirect standardisation, it is inappropriate to compare sub-geographies with each other – all comparisons should be made with the national figure.)

Denominator

The number of finished continuous inpatient (CIP) spells within selected medical and surgical specialties with a discharge date up to 31st March in the financial year of analysis.

Numerator

The number of finished and unfinished CIP spells intersecting the respective financial year, plus those up to 30 days into the next financial year that are emergency admissions within 0-29 days (inclusive) of the last, previous discharge from hospital.

NHS Digital reference code

The records for the denominator and the numerator are taken from Hospital Episode Statistics Admitted Patient Care (HES APC) data linked as Continuous Inpatient (CIP) Spells.

This section sets out how this indicator is constructed.

Denominator

The number of finished CIP spells with a valid CCG value, within selected medical and surgical specialties, with a discharge date up to 31st March within the financial year of analysis.  Day cases, spells with a discharge coded as death, maternity spells (based on specialty, admission or discharge episode type and primary diagnosis), and those with mention of a diagnosis of cancer or chemotherapy for cancer anywhere in the spell are excluded.  Patients with a mention of a diagnosis of cancer or chemotherapy for cancer anywhere in the 365 days prior to admission are also excluded.

The following fields and values are used to filter the denominator.

The CIP Spell has:

DISDATE (discharge date)

Filters the whole period of the financial year being reported on., e.g. between 01/04/2017 and 31/03/2018 inclusive AND is equal to or after the ADMIDATE (date of admission).

DISMETH (method of discharge)

1 or 3 (Discharged on clinical advice or discharged by a mental health review tribunal, i.e. not died, self-discharged or a baby or still in hospital).

ADMIMETH (admission method)

Contains: 11, 12, 13, 21, 22, 23, 24, 25, 28, 2A, 2B, 2C, 2D, 31, 32, 81, 82, 83, 84 or 89.

The FIRST episode in the CIP spell has:

EPIORDER (episode order)

Equals 1.

CCG_RESPONSIBILITY

Valid value i.e. a valid CCG code for the reporting year (excludes unknowns).  Note: legacy organisations are mapped to new organisation codes when known.

EPITYPE (episode type)

Equals 1.  General episode types only.

CLASSPAT (patient classification)

Equals 1.  Ordinary Admissions only.

STARTAGE (age at start of episode)

Value between 0 and 120 and 7001 and 7007 only.  Value between 7001 and 7007 are converted to 0.

DOB (patient date of birth)

Not 01/01/1900 or 01/01/1901 representing unknown.

SEX (sex of patient)

Equals 1 or 2.

AND the LAST episode in the CIP spell has:

EPITYPE (episode type)

Equals 1.  General episode types only.

AND exclude the whole CIP spell where ANY episode in the spell has:

DIAG_01 - DIAG_20 (any diagnosis)

C00*-C97*, D37*-D48* (any mention of a diagnosis of cancer) OR Z51.1* (any mention of chemotherapy for cancer).  Applies both within the reporting year or anywhere in the 365 days prior to admission for the patient.

OR

TRETSPEF (the specialised service within which the patient is treated)

Not 501, 560, 610.

DIAG_01 (primary diagnosis)

Does not begin with ‘O’ (Obstetrics).

There is an additional 3 step piece of logic which is designed to ensure that the spells are allocated to the most appropriate diagnosis/procedure group for standardisation:

Step A

Searching all episodes from first to last, identify spells where there is a valid procedure (i.e. not being either: (1) a null value, (2) "-" meaning “no operation performed”, (3) "&" meaning “not known”.) and surgical specialty (taken from the episode where the procedure was found).  These spells are standardised by procedure subgroup (procedures beginning with ‘Y’ or ‘Z’ are standardised in the ‘no procedure’ basket).

Step B

Excluding spells selected in step A, select spells where treatment function specialty in the first episode is surgical, these spells are standardised under the ‘no procedures’ basket.

Step C

Excluding those spells selected in step A and step B, select spells where treatment function specialty of the first episode is medical.  These spells are standardised by diagnosis subgroup.

In all cases the first three characters of the respective diagnosis or procedure code is used.

Lists of specialties and sub-groups used above for filtering/standardisation are:

Medical Specialties: 190, 192, 223, 242, 251, 252, 253, 254, 255, 256, 257, 258, 259, 260, 261, 262, 263, 264, 300, 301, 302, 303, 304, 305, 306, 307, 308, 309, 310, 311, 313, 314, 315, 316, 317, 318, 319, 320, 321, 322, 323, 324, 325, 327, 328, 329, 330, 331, 340, 341, 342, 343, 344, 345, 346, 350, 352, 360, 361, 370, 371, 400, 401, 410, 420, 421, 422, 424, 430, 450, 460, 501, 502, 503, 510, 520, 560, 610, 620.

Surgical Specialties: 100, 101, 102, 103, 104, 105, 106, 107, 108, 110, 120, 130, 140, 141, 142, 143, 144, 150, 160, 161, 170, 171, 172, 173, 174, 180, 191, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 241.

Numerator

The number of finished and unfinished CIP spells that are emergency admissions within 0-29 days (inclusive) of the last, previous discharge from hospital (see denominator), including those where the patient dies, but excluding the following: those with a main specialty and primary diagnosis upon readmission coded under obstetrics and those where the readmission spell has a diagnosis of cancer (other than benign or in situ) or chemotherapy for cancer coded anywhere in the spell.

The date of the last, previous discharge from hospital, and the date and method of admission from the following CIP spell, are used to determine the interval between discharge and emergency readmission.

The numerator is based on a pair of spells, the discharge spell and the next subsequent readmission spell (this spell must meet the numerator criteria).  The selection process thus carries over the characteristics of the denominator for the discharge spell and applies additional ones to the readmission spell.

The following fields and values are used for the numerator.

The CIP spell has:

ADMIDATE (date of admission)

Between 1 April of the financial year reporting on and 30 April of the following year.  Filters the whole period of the financial year being reported on plus one month., e.g. between 01/04/2017 and 30/04/2018 inclusive.

ADMIMETH (admission method)

Contains: 21, 22, 23, 24, 25, 28, 2A, 2B, 2C, or 2D (emergency admissions).

The FIRST episode in the CIP spell has:

EPIORDER (episode order)

Equals 1.

CCG_RESPONSIBILITY

Valid value i.e. a known CCG code for the reporting year, i.e. excludes unknowns.  Note: legacy organisations are mapped to new organisation codes when known.

EPITYPE (episode type)

Equals 1.  General episode types only.

CLASSPAT (patient classification)

Equals 1.  Ordinary Admissions only.

AND ADMIDATE from the FIRST episode of the Readmission CIP spell minus DISDATE from the LAST episode in Discharge CIP spell ≤ 29 days.  Note: where there is more than one readmission within 30 days, each readmission is counted once, in relation to the previous discharge.

AND exclude where ANY episode in the CIP spell has:

DIAG_01 - DIAG_20 (any diagnosis)

C00*-C97*, D37*-D48* (any mention of a diagnosis of cancer) OR Z51.1* (any mention of chemotherapy for cancer).  Applies both within the reporting year or anywhere in the 365 days prior to admission for the patient.

OR

TRETSPEF (the specialised service within which the patient is treated)

Not 501, 560, 610.

DIAG_01 (primary diagnosis)

Does not begin with ‘O’ (Obstetrics).

The first step is to calculate the casemix-specific rates for the standard population.  The dataset described above is aggregated with counts of readmissions (numerator) and previous discharges (denominator) grouped by:

• Age based on STARTAGE using the following age bands: <1, 1-4, 5-9, 10-15,16-64, 65-74, 75-84, 85+
• Sex based on SEX as 1, 2 (male and female)
• Method of admission of discharge spell based on ADMIMETH as elective or non-elective
• Medical or Surgical specialties based on TRETSPEF.

Then either:

• Diagnosis group (within medical specialties) based on DIAG_01 (primary diagnosis) and expressed to 3 characters

OR

• Procedure group (within surgical specialties) based on OPERTN_01 (primary procedure) expressed to 3 characters.

For each group in this casemix, the casemix specific readmission rates are calculated and are then applied to the casemix structure of the subject population i.e. each CCG.  This gives an expected number of events against which the observed number of events may be compared.

Indirect standardisation involves the calculation of the ratio of observed number of events (for the CCG) and the number of events that would be expected if it had experienced the same event rates as those of patients in England, given the casemix of age, sex, method of admission and diagnosis / procedure of its patients.

\({ IS_{Ratio} = \frac {O}{E} \times 100 = \frac {\sum_i O_i}{\sum_i E_i} \times 100 = \frac {\sum_i O_i}{\sum_i n_i \lambda_i} \times 100 }\)

(expressed per 100 denominator population)

where:

\(\mathsf{O_i} \) is the observed number of events in the subject population in casemix group \(\mathsf{i} \)

\(\mathsf{E_i} \) is the expected number of events in the subject population in casemix group \(\mathsf{i} \)

\(\mathsf{n_i} \) is the number of individuals in the subject population in casemix group \(\mathsf{i} \)

\(\mathsf{\lambda_i} \) is the crude age-specific rate in the standard population in casemix group \(\mathsf{i} \).

This standardised ratio is then converted into a rate by multiplying it by the overall event rate of patients in England.

\({ IS_{Rate} = \frac {O}{E} \times \lambda \times 100 = \frac {\sum_i O_i}{\sum_i E_i} \times \lambda \times 100 = \frac {\sum_i O_i}{\sum_i n_i \lambda_i} \times \lambda \times 100 }\)

(expressed per 100 denominator population)

where:

\(\mathsf{O_i} \) is the observed number of events in the subject population in casemix group \(\mathsf{i} \)

\(\mathsf{E_i} \) is the expected number of events in the subject population in casemix group \(\mathsf{i} \)

\(\mathsf{n_i} \) is the number of individuals in the subject population in casemix group \(\mathsf{i} \)

\(\mathsf{\lambda_i} \) is the crude age-specific rate in the standard population in casemix group \(\mathsf{i} \)

\(\mathsf{\lambda} \) is the overall crude rate in the standard population.

The lower and upper limits of the 95% confidence interval for the indirectly standardised rate are calculated by finding the lower and upper limits of the standardised ratio and multiplying by the overall crude rate of the standard population.  The Byar’s approximation is used as it is a sufficiently accurate approximation to the Poisson probabilities (from Breslow NE and Day NE. Statistical Methods in Cancer Research, Volume II: The Design and Analysis of Cohort Studies. Lyon: International Agency for Research on Cancer, World Health Organization, 1987: 69).

The 95% limits are given by:

\({ IS_{Rate(LL)} = \frac {O}{E} \times \Bigg \{ 1 - \frac {1}{9O} - \frac {1.96}{\sqrt[3]{(O+1)}} \Bigg \}^3 \times \lambda \times 100 }\)

\({ IS_{Rate(UL)} = \frac {(O+1)}{E} \times \Bigg \{ 1 - \frac {1}{9(O+1)} + \frac {1.96}{\sqrt[3]{(O+1)}} \Bigg \}^3 \times \lambda \times 100 } \)

(expressed per 100 denominator population)

where:

\(\mathsf{O} \) is the total observed number of events in the subject population

\(\mathsf{E} \) is the total expected number of events in the subject population

\(\mathsf{\lambda} \) is the overall crude rate in the standard population.

When output to a table, such as excel or CSV, the following is produced.