Hospital guide methodology

The Dr Foster Hospital Guide provides information on NHS acute hospitals with more than 300 beds, some smaller NHS acute hospitals that provide key services to a geographical area and selected specialist hospitals.

Acute and Specialist hospitals are run by NHS Trusts in England and Wales, Health and Social Services Trusts in Northern Ireland, and by NHS Boards in Scotland. A trust or board may run one or more hospitals depending on its location. The information in this guide can be at trust/board level or hospital level, depending on how it has been collected.

About the indicators

There are a number of different clinical indicators in the hospital guide presented for a range of clinical procedures. This section provides more detailed information on how the various indicators are calculated, where the data comes from and what time periods are covered.

Hospital homepages

General hospital information

Metric

Information about the facilities available at individual hospitals has been drawn from the Hospital Estates and Facilities Statistics data-set. The fields used include:

number of beds
percentage of single rooms
total parking spaces available for patients/visitors
average parking fee per hour

More information can be found on the Hospital Estates and Facilities Statistics website.

Data Source

The Information Centre - Hospital Estates and Facilities Statistics

Time frame

April 2011 - March 2012

Basis

Hospital Level

Procedural Scorecard Indicators

Length of Stay

Metric

Average number of days spent in an acute trust for planned treatment.

Numerator

The total number of days in elective inpatient spells, at an acute trust for a given main procedure.

The difference between the date of discharge and date of admission is used to determine the interval of stay within the trust (days after possible transfers from the trust are not included). Day cases are omitted from the calculation (a day case is an inpatient spell where the patient is not intended and does remain overnight).

Denominator

A count of the number of elective spells occurring at a trust (NHS) or a site (IS) during the current extract period. A count of the number of elective spells occurring at a provider during the extract period.

Data Source

Commissioning Data Sets (CDS)

Time frame

July 2011 - June 2012

Basis

Acute Trust

Statistical methods used

The indicator is a statistical mean (average) of the raw days spent in hospital: the total number of days accruing from finished spells, divided by the number of spells. It is not standardised.

National median

A national median length of stay for each relevant procedure is calculated by arranging the length of stay for all relevant spells (elective spells in NHS acute hospitals in England) in ascending order to identify the middle value; where there are an even number of values, an average of the two central values is taken.

Relative risk of Readmission

Metric

A ratio of the observed number of emergency readmissions at a given trust to the expected number of 28-day readmissions for a particular procedure.

Numerator

The total number of emergency readmissions resulting from inpatient spells at an acute trust for a given procedure. The emergency readmission must occur within a 28-day period - 0-28 days inclusive - from the discharge date of the originating spell.

Denominator

The expected number of emergency readmissions resulting from every finished inpatient spell at an acute trust for a given procedure where the procedure is the main procedure in the spell.

The expected number of readmissions is obtained by modelling the risk of readmission.

Each inpatient spell has a risk of readmission associated with it based upon deprivation, sex, age, co-morbidity, the number of emergency admissions a patient has had in the previous 12 months and the presence of a palliative care episode. The risks are summed to obtain the overall expected number of readmissions.

Data Source

Commissioning Data Sets (CDS)

Time frame

July 2011 - June 2012

Basis

Acute Trust

Statistical methods used

Logistic regression models of readmissions are built from all years of HES data from 1996/7 onwards based upon the available case-mix factors. The factors adjusted for are: age, sex and method of admission (non-elective or elective), year of admission, deprivation quintile, the Charlson index of comorbidity (see for example, Sundararajan et al, Journal of Clinical Epidemiology 2004; 57: 1288-1294, available via ScienceDirect), the number of emergency admissions a patient has had in the previous 12 months and the presence of a palliative care episode. Based upon these parameters a logistic regression model is developed which predicts the risk of readmission for an inpatient spell.

A 97.5% confidence interval is calculated for the indicator using Byar's approximation. The confidence interval enables a banding of the indicator for comparison purposes.

Banding

Relative risk is not a quantitative benchmarking measure. It is not intended to give exact readmission rates for a procedure, it is designed to highlight and show where a trust's readmission is not in line with outcomes nationwide.

There are three reported bands:

if the lower limit of the confidence interval is greater then 100 than the assigned band is red, "more than expected"
if the upper limit of the confidence interval is less then 100 than the assigned band is green, "less than expected"
otherwise, the assigned band is blue, "as expected"

1 Year Hospital Standardised Mortality Ratio

Metric

The ratio of the observed number of in-hospital deaths during admissions with a Hospital Standardised Mortality Ratio (HSMR) diagnosis to the expected number of deaths, multiplied by 100

Observed

Denominator superspells with method of discharge as death (DISMETH=4,5)

Denominator

Superspells containing a spell with a primary dominant diagnosis of any of the 56 CCS groups that comprise the HSMR basket

See appendix M: HSMR basket

Data Source

SUS - CDS

Time frame

July 2011 - June 2012

Basis

Acute Trust

Statistical methods

Logistic Regression

Expected number of in-hospitals deaths is derived from logistic regression, adjusting for factors to indirectly standardise for differences in case-mix.

Adjustments are made for:

Sex
Age on admission (in five year bands up to 90+)
Interactions between age on admission (in five year bands up to 90+) and charlson co-morbidity score**
Admission method (non-elective or elective)
Socio-economic deprivation quintile of the area of residence of the patient (based on the Carstairs Index)
Diagnosis/procedure subgroup
Co-morbidities (based on Charlson score)
Number of previous emergency admissions
Year of discharge (financial year)
Palliative care (if any episode in the spell has the treatment function code 315 or contains ICD10 code Z515 in any of the diagnoses fields)
Month of admission
Source of admission

**new to logistic regression model in 2011

Relative Risk

The ratio is calculated by dividing the actual number of deaths by the expected number and multiplying the figure by 100. It is expressed as a relative risk, where a risk rating of 100 represents the national average. If the trust has an SMR of 100, that means that the number of patients who died is exactly as it would be expected taking into account the standardisation factors. An SMR above 100 means more patients died than would be expected; one below 100 means that fewer than expected died.

Control Limits

Control limits tell us the range of values which are consistent with random or chance variation. Data points falling within the control limits are consistent with random or chance variation and are said to display 'common-cause variation'; for data points falling outside the control limits, chance is an unlikely explanation and hence they are said to display 'special-cause variation' - that is, where the trust's rate diverges significantly from the national rate.

Data points falling above the upper 99.8% poisson control limit are said to be significantly 'higher than expected', data points falling below the lower 99.8% poisson control limit are said to be significantly 'lower than expected', otherwise 'within expected range'.

Notes

The HSMR basket of CCS groups accounts for approximately 83% of all in-hospital deaths in England.

See 'HSMR Toolkit' for full methodological detail

Superspell: a group of spells linked by transfer

3 Year Hospital Standardised Mortality Ratio

Metric

The ratio of the observed number of in-hospital deaths during admissions with a Hospital Standardised Mortality Ratio (HSMR) diagnosis to the expected number of deaths, multiplied by 100

Observed

Denominator superspells with method of discharge as death (DISMETH=4,5)

Denominator

Superspells containing a spell with a primary dominant diagnosis of any of the 56 CCS groups that comprise the HSMR basket

See appendix M: HSMR basket

Data Source

SUS - CDS

Time frame

Financial years 2009/10, 2011/10, 2011/12

Basis

Acute Trust

Statistical methods

Logistic Regression

Expected number of in-hospitals deaths is derived from logistic regression, adjusting for factors to indirectly standardise for differences in case-mix.

Adjustments are made for:

Sex
Age on admission (in five year bands up to 90+)
Interactions between age on admission (in five year bands up to 90+) and charlson co-morbidity score**
Admission method (non-elective or elective)
Socio-economic deprivation quintile of the area of residence of the patient (based on the Carstairs Index)
Diagnosis/procedure subgroup
Co-morbidities (based on Charlson score)
Number of previous emergency admissions
Year of discharge (financial year)
Palliative care (if any episode in the spell has the treatment function code 315 or contains ICD10 code Z515 in any of the diagnoses fields)
Month of admission
Source of admission

**new to logistic regression model in 2011

Relative Risk

Control Limits

Notes

The HSMR basket of CCS groups accounts for approximately 83% of all in-hospital deaths in England.

See 'HSMR Toolkit' for full methodological detail

Superspell: a group of spells linked by transfer

One-year revision rates for primary hip and knee replacements

Metric

The proportion of joint replacements with a revision procedure within 365 days of the initial (index) procedure, over the total number of joint replacements carried out at the trust over a three year period.

Three years of index procedures are combined to provide sufficient numbers at trust level. A further year of data is needed to allow a year's follow-up for every index procedure. Index operations in 2004/5 to 2006/7 give rise, potentially, to revisions between 2004/5 and 2007/8.

Note that only one revision within 365 days per patient is counted (some people can have several), and revisions are matched to side of index procedure (right or left).

Construction

Index procedure codes for primary total hip replacement (THR)

Codes: oper1 in ('W371','W381','W391') or (oper1='W581' and (oper2='Z843' or oper3='Z843' or oper4='Z843'))

Description: These cover primary total replacement including resurfacing procedures. For the latter, the Z code is required in any secondary op field (oper2-oper12 - only 2 to 4 shown above for brevity) to specify hip.

Revision procedure codes for THR

Codes: oper1 in ('W373','W383','W393', 'W372','W374', 'W382','W384','W392','W394') or (oper1='W582' and (oper2='Z843' or oper3='Z843' or oper4='Z843')) or oper2 in ('W580', 'W370', 'W380', 'W390') or oper3 in ('W580', 'W370', 'W380', 'W390') or oper4 in ('W580', 'W370', 'W380', 'W390');

Description: These cover revisions (primary op field), conversions (in any secondary op field) and "attention to joint" (primary op field).

Index procedure codes for primary total knee replacement (TKR)

Codes: oper1 in ('W401','W411','W421') or (oper1='W581' and (oper2='Z846' or oper3='Z846' or oper4='Z846'))

Description: These cover primary total replacement including resurfacing procedures. For the latter, the Z code is required in any secondary op field to specify knee.

Revision procedure codes for TKR

Codes: oper1 in ('W403','W413','W423', 'W402','W404', 'W412','W414', 'W422','W424') or (oper1 in ('W580','W582') and (oper2='Z846' or oper3='Z846' or oper4='Z846')) or oper2 in ('W400', 'W410', 'W420') or oper3 in ('W400', 'W410', 'W420') or oper4 in ('W400', 'W410', 'W420')

Description: These cover revisions (primary op field), conversions (in any secondary op field) and "attention to joint" (primary op field).

Data Source

Commissioning Data Sets (CDS)

Time frame

April 2007 - March 2011

Basis

Acute Trust where index procedure took place.

Statistical methods used

Crude rate

Banding

Upper and lower control limits are calculated at 97.5% level of significance and these are used to band the crude rate.

There are three reported bands:

if the crude rate of revisions is greater than the upper control limit then the assigned band is red, "more than expected"
if the crude rate of revisions is less than the lower control limit then the assigned band is green, "less than expected"
otherwise, the assigned band is blue, "as expected"

Waiting time - Inpatient

Metric: By procedure, for elective care, the median wait time (elapsed number of days) between the decision to admit and the day of admission
Numerator: Not applicable.
Denominator: Not applicable.
Data Source: Commissioning Data Sets (CDS)
Time frame: July 2011 - June 2012
Basis: Acute Trust
Statistical methods used: The median is a statistical term that identifies the middle observation of a data set ordered in increasing value; where there is an even number of values, an average of the two central values is taken.
National median: An identical calculation the basis of which is national data (all English NHS Acute Trusts).

Waiting time - Outpatient

Metric

By specialty, the median wait time (elapsed number of days) for the first outpatient attendance following a GP referral.

In order to determine what specialty to show in relation to a procedure, we identify, nationally, all the discharge specialties that a procedure has been assigned to. We then identify the discharge specialty with the greatest number of spells assigned to it.

Numerator

Not applicable.

Denominator

Not applicable.

Data Source

Commissioning Data Sets (CDS)

Time frame

July 2011 - June 2012

Basis

Acute Trust

Statistical methods used

The median is a statistical term that identifies the middle observation of a data set ordered in increasing value; where there is an even number of values, an average of the two central values is taken.

National median

An identical calculation the basis of which is national data (all English NHS Acute Trusts).

Relative risk of inpatient admission

Metric

By procedure, a ratio of the observed number of inpatient admissions to the total number of all elective admissions (inpatient and day cases).

Numerator

By procedure, the total number of observed inpatient admissions.

Denominator

The expected number of inpatient admissions.

The expected number of admissions is obtained by modelling the risk of admission.

Each inpatient spell has a risk of admission associated with it based upon deprivation, sex, age, co-morbidity, the number of emergency admissions a patient has had in the previous 12 months and the presence of a palliative care episode. The risks are summed to obtain the overall expected number of inpatient admissions.

Data Source

Commissioning Data Sets (CDS)

Time frame

July 2011 - June 2012

Basis

Acute Trust

Statistical methods used

Logistic regression models of inpatient admissions are built from all years of HES data from 1996/7 onwards based upon the available case-mix factors. The factors adjusted for are: age, sex and method of admission (non-elective or elective), year of admission, deprivation quintile, the Charlson index of comorbidity (see for example, Sundararajan et al, Journal of Clinical Epidemiology 2004; 57: 1288-1294, available via ScienceDirect), the number of emergency admissions a patient has had in the previous 12 months and the presence of a palliative care episode. Based upon these parameters a logistic regression model is developed which predicts the risk of admission for an inpatient spell.

A 97.5% confidence interval is calculated for the indicator using Byar's approximation. The confidence interval enables a banding of the indicator for comparison purposes.

Banding

Relative risk is not a quantitative benchmarking measure. It is not intended to give exact admission rates for a procedure, it is designed to highlight where a trust's inpatient admissions are not in line with outcomes nationwide.

There are three reported bands:

if the lower limit of the confidence interval is greater than 100 then the assigned band is red, "more than expected"
if the upper limit of the confidence interval is less than 100 then the assigned band is green, "less than expected"
otherwise, the assigned band is blue, "as expected"

Mortality rate: Hip replacement

Metric

Relative risk of mortality within 30 days of a procedure, that is, the ratio of the observed number of deaths at a given trust to the expected number of 30-day mortalities for a particular procedure (this ratio is multiplied by 100).

Numerator

The total number of observed mortalities (in-hospital 30 days, 0-30 days inclusive) resulting from every finished inpatient spell at an acute trust for an elective Hip replacement procedure (the numerator is multiplied by 100).

The mortality must occur within 30 days of the date of the main procedure in the spell.

Denominator

The expected number of mortalities resulting from every finished inpatient spell at an acute trust for an elective Hip replacement.

The expected number of mortalities is obtained by modelling the risk of mortality using a logistic regression model. Each inpatient spell has a risk of mortality associated with it based upon deprivation, sex, age, co-morbidity. The risks are summed to obtain the overall expected number of mortalities.

Data Source

Commissioning Data Sets (CDS)

Time frame

April 2008 - March 2011

Basis

Acute Trust

Statistical methods used

Logistic regression models of the risk of post-operative mortality are built from all years of HES data from 1996/7 onwards based upon the available case-mix factors. The major factors adjusted for are: age, sex and method of admission (non-elective or elective), year of admission and deprivation quintile. Three further variables are included in the models: the Charlson index of comorbidity, the number of emergency admissions a patient has had in the previous 12 months and the presence of a palliative care episode. Based upon these parameters a logistic regression model is developed which predicts the risk of mortality for an inpatient spell. Since the admission events to a provider are (assumed) independent these risks are summed at the provider level for NHS Trusts, to yield the expected number of port operative (30 day) deaths. 99.8% (Poisson) control limits is calculated for the indicator. The control limits enable a banding of the indicator for comparison purposes.

Banding

Any relative risk of mortality is not a quantitative benchmarking measure. It is not intended to give exact mortality rates for a procedure, it is designed to highlight and show where a trust's mortality rate is not in line with outcomes nationwide.

There are four bands:

Where the relative risk of mortality is higher than the upper control limit the assigned band is red, "More than expected"
Where the relative risk of mortality is below the lower control limit the assigned band is green, "Less than expected"
Where the relative risk of mortality falls between the upper and lower control limits the assigned band is blue, "As expected"
If the expected figure is less than 1, it is likely that the sample size is too small to allow any meaningful comparison and the assigned band is black, "Sample too small"

Note: Where the expected figure lies between 1 and 5, the banding is restricted to red, "More than expected", and blue, "As expected".

Mortality rate: Knee replacement

Metric

Numerator

The total number of observed mortalities (in-hospital 30 days, 0-30 days inclusive) resulting from every finished inpatient spell at an acute trust for an elective Knee replacement procedure (the numerator is multiplied by 100).

The mortality must occur within 30 days of the date of the main procedure in the spell.

Denominator

The expected number of mortalities resulting from every finished inpatient spell at an acute trust for an elective knee replacement.

Data Source

Commissioning Data Sets (CDS)

Time frame

April 2008 - March 2011

Basis

Acute Trust

Statistical methods used

Banding

There are four bands:

Where the relative risk of mortality is higher than the upper control limit the assigned band is red, "More than expected"
Where the relative risk of mortality is below the lower control limit the assigned band is green, "Less than expected"
Where the relative risk of mortality falls between the upper and lower control limits the assigned band is blue, "As expected"
If the expected figure is less than 1, it is likely that the sample size is too small to allow any meaningful comparison and the assigned band is black, "Sample too small"

Note: Where the expected figure lies between 1 and 5, the banding is restricted to red, "More than expected", and blue, "As expected".

Mortality rate: Abdominal Aortic Aneurysm (AAA) repair

Metric

Numerator

The total number of observed mortalities (in-hospital 30 days, 0-30 days inclusive) resulting from every finished inpatient spell at an acute trust for an elective Abdominal Aortic Aneurysm (AAA) repair (the numerator is multiplied by 100).

The mortality must occur within 30 days of the date of the main procedure in the spell.

Denominator

The expected number of mortalities resulting from every finished inpatient spell at an acute trust for an elective Abdominal Aortic Aneurysm (AAA) repair.

Data Source

Commissioning Data Sets (CDS)

Time frame

April 2008 - March 2011

Basis

Acute Trust

Statistical methods used

Banding

There are four bands:

Where the relative risk of mortality is higher than the upper control limit the assigned band is red, "More than expected"
Where the relative risk of mortality is below the lower control limit the assigned band is green, "Less than expected"
Where the relative risk of mortality falls between the upper and lower control limits the assigned band is blue, "As expected"
If the expected figure is less than 1, it is likely that the sample size is too small to allow any meaningful comparison and the assigned band is black, "Sample too small"

Note: Where the expected figure lies between 1 and 5, the banding is restricted to red, "More than expected", and blue, "As expected".

If the expected figure lies between 1 and 5, the banding is restricted to Bands 1 and 2. When the expected figure is greater than five, Bands 1, 2 and 3 pertain.

MRSA rates

Metric

This indicator shows the Trust apportioned rate of blood infections caused by MRSA for all patients. This includes those who are admitted as an emergency and those who are elective patients i.e. they have planned their treatment in advance. The data is presented as the number of cases per 100,000 bed days. A bed day is defined as one person in hospital for one night.

You can find out more information by visiting the Health Protection Agency website.

Data Source

The Health Protection Agency

Time frame

Financial year 2011/2012

Basis

Trust level information

Clostridium difficile infection rates

Metric

This indicator shows the rate per 100,000 bed-days for specimens taken from patients aged 2 years and over (Trust apportioned cases. A bed day is defined as one person in hospital for one night).

You can find out more information by visiting the Health Protection Agency website.

Data Source

The Health Protection Agency

Time frame

Financial year 2011/2012

Basis

Trust level information

NHS Trust homepage Indicators

Mortality

1 Year Hospital Standardised Mortality Ratio

Metric

The ratio of the observed number of in-hospital deaths during admissions with a Hospital Standardised Mortality Ratio (HSMR) diagnosis to the expected number of deaths, multiplied by 100

Observed

Denominator superspells with method of discharge as death (DISMETH=4,5)

Denominator

Superspells containing a spell with a primary dominant diagnosis of any of the 56 CCS groups that comprise the HSMR basket

See appendix M: HSMR basket

Data Source

SUS - CDS

Time frame

July 2011 - June 2012

Basis

Acute Trust

Statistical methods

Logistic Regression

Expected number of in-hospitals deaths is derived from logistic regression, adjusting for factors to indirectly standardise for differences in case-mix.

Adjustments are made for:

Sex
Age on admission (in five year bands up to 90+)
Interactions between age on admission (in five year bands up to 90+) and charlson co-morbidity score**
Admission method (non-elective or elective)
Socio-economic deprivation quintile of the area of residence of the patient (based on the Carstairs Index)
Diagnosis/procedure subgroup
Co-morbidities (based on Charlson score)
Number of previous emergency admissions
Year of discharge (financial year)
Palliative care (if any episode in the spell has the treatment function code 315 or contains ICD10 code Z515 in any of the diagnoses fields)
Month of admission
Source of admission

**new to logistic regression model in 2011

Relative Risk

Control Limits

Notes

The HSMR basket of CCS groups accounts for approximately 83% of all in-hospital deaths in England.

See 'HSMR Toolkit' for full methodological detail

Superspell: a group of spells linked by transfer

3 Year Hospital Standardised Mortality Ratio

Metric

The ratio of the observed number of in-hospital deaths during admissions with a Hospital Standardised Mortality Ratio (HSMR) diagnosis to the expected number of deaths, multiplied by 100

Observed

Denominator superspells with method of discharge as death (DISMETH=4,5)

Denominator

Superspells containing a spell with a primary dominant diagnosis of any of the 56 CCS groups that comprise the HSMR basket

See appendix M: HSMR basket

Data Source

SUS - CDS

Time frame

Financial years 2009/10, 2010/11, 2011/12

Basis

Acute Trust

Statistical methods

Logistic Regression

Expected number of in-hospitals deaths is derived from logistic regression, adjusting for factors to indirectly standardise for differences in case-mix.

Adjustments are made for:

Sex
Age on admission (in five year bands up to 90+)
Interactions between age on admission (in five year bands up to 90+) and charlson co-morbidity score**
Admission method (non-elective or elective)
Socio-economic deprivation quintile of the area of residence of the patient (based on the Carstairs Index)
Diagnosis/procedure subgroup
Co-morbidities (based on Charlson score)
Number of previous emergency admissions
Year of discharge (financial year)
Palliative care (if any episode in the spell has the treatment function code 315 or contains ICD10 code Z515 in any of the diagnoses fields)
Month of admission
Source of admission