SPARRA Frequently Asked Questions
Q1. What does SPARRA stand for?
Q3. How are SPARRA scores calculated?
Q4. What factors are used to calculate SPARRA scores?
Q5. Who are SPARRA scores calculated for?
Q6. What do SPARRA scores look like?
Q7. What can SPARRA scores be used for?
Q8. How is the performance of the SPARRA model evaluated?
Q9. How are SPARRA data distributed?
Q10. How frequently are SPARRA scores updated?
Q11. Who currently receives SPARRA lists?
Q12. How can I receive SPARRA data for my area?
Q13. Who is the relevant signatory for my confidentiality form?
Q14. Who should I contact for further information?
Q1. What does SPARRA stand for?
Scottish Patients at Risk of Readmission and Admission
Q2. What is SPARRA?
Scottish Patients at Risk of Readmission and Admission (SPARRA) is an algorithm developed by Information Services Division (ISD) to predict a patient's risk of being admitted to hospital as an emergency in a particular year.
Q3. How are SPARRA scores calculated?
To build the SPARRA algorithm, the statistical technique of logistic regression was used. By examining patient characteristics, and which patients were admitted as an emergency in a particular outcome year, characteristics that can be used as predictors of emergency admission were identified. The extent to which these characteristics are present for each patient in a cohort can then be used to estimate their risk of emergency hospitalisation. The current SPARRA algorithm, SPARRA Version 3 (which was implemented in January 2012) links patient level data relating to hospital admissions, prescriptions (dispensed items), new outpatient attendances, Emergency Department attendances and psychiatric hospital admissions, and employs characteristics available from the combined dataset to calculate a patient's risk score. A key feature of SPARRA Version 3 is the division of the SPARRA cohort into three sub-cohorts: frail elderly, long term conditions, and younger Emergency Department. These sub-cohorts each have their own specific set of risk characteristics tailored to that particular population. Please see the SPARRA Version 3 web page for further detail.
Scores are currently calculated at quarterly intervals for patients for the forthcoming year, on the basis of the previous three years of patient history. The period of history varies depending on the data source. For example, in the case of hospitalisation data the most recent three years available data is used, whereas for Emergency Department attendances it is the most recent twelve months of information. By way of illustration, in relation to the January 2012 output, the three year patient history of hospital admission used was January 2009 to December 2011, the twelve month patient history of Emergency Department attendance used January 2011 to December 2011, and the risk score calculated was for the time period January 2012 to December 2012.
Q4. What factors are used to calculate SPARRA scores?
The SPARRA algorithm takes a number of factors into account in assessing a patient's risk of emergency admission. Version 3 SPARRA includes the following factors to estimate an individual's risk:
- Numbers of emergency, elective and daycase admissions in the last three years
- Numbers of emergency and elective bed days in the last three years
- Emergency admissions with alcohol or substance misuse related diagnoses in the last three years
- Number of long term conditions identified by the diagnoses relating to previous hospital admissions (1981 onwards)
- Number of prescriptions (dispensed items) in the last twelve months for selected BNF Chapters and/or BNF Sections
- Number of different BNF Sections for which the patient has prescriptions (dispensed items) in the last twelve months
- Number of admissions to psychiatric hospital in the last three years
- Number of Emergency Department attendances in the previous twelve months
- Number of new outpatient attendances in the previous twelve months
- Number of new outpatient attendances to mental health specialties in the previous twelve months
- Evidence, through hospitalisation and/or prescribing history, of Parkinson's Disease, MS, epilepsy or dementia.
- Age group at last emergency admission
- Deprivation - Scottish Index of Multiple Deprivation (SIMD) Decile.
The precise combination of factors and definition applied varies between the three SPARRA sub-cohorts (frail eldery, long term condition, and younger Emergency Department). Please see the SPARRA Version 3 web page for further detail.
Q5. Who are SPARRA scores calculated for?
SPARRA Version 3 currently calculates risk scores for approximately 3.2 million individuals who have a recent history of hospitalisation, prescribing, Emergency Department attendance, new outpatient attendance or psychiatric hospital admission. Further improvements in data quality and coverage for the newer datasets used by SPARRA (prescribing and Emergency Department) are expected to increase the SPARRA cohort to about 3.5 million.
Q6. What do SPARRA scores look like?
SPARRA scores can range from 1 to 99% for patients in the cohort. Patients with a score of 50%, for example, are generally said to have a 1 in 2 chance of being admitted to hospital in the prediction year. Recipients can opt to receive data for patients at risk thresholds subject to certain minima (see Q9. How are SPARRA data distributed?).
Q7. What can SPARRA scores be used for?
There are two main areas in which this information is beneficial; case finding and service planning.
- Case Finding: In recent years there has been growing acceptance of the need to shift from a health care system that is geared towards reactive hospital-based treatment of acute conditions to one which is founded on a preventative, anticipatory approach. In order to embrace such a system of anticipatory care, it is essential to have a methodologically sound technique which enables stratification of the population. SPARRA allows health care professionals to identify patients with complex care needs who are likely to benefit most from anticipatory health care. They can ensure that patients on the SPARRA list are known to the relevant professionals and that the appropriate level of care is being provided. Such identification could be pivotal in identifying patients at an earlier stage in the development of their condition and in preventing future emergency hospitalisation.
- Service Planning: SPARRA can also be used to locate groups of patients who would benefit from specific interventions or services. This relates to anticipatory care in a broad sense but is specifically concerned with maximising the value or impact of services. Again, SPARRA can play a key role in this by providing data which stratifies the population by risk and identifies areas of need. This is essential in order to make rational and coordinated decisions about health service planning.
Q8. How is the performance of the SPARRA model evaluated?
- Positive Predictive Value (PPV) is the proportion of patients above a specified risk threshold who were admitted in the prediction year. It can be regarded as a measure of accuracy i.e. the proportion of patients accurately identified as being at risk of emergency hospitalisation. Using the current SPARRA algorithm the PPV observed for those with a risk score of 50+% is approximately 60%. This means that typically 60% of patients above this risk threshold will go on to have an emergency admission in the prediction year.
- Sensitivity is the number of patients above a specified risk threshold who were admitted in the prediction year as a proportion of all patients in the SPARRA cohort who were admitted in the prediction year. It can be regarded as a measure of coverage i.e. the proportion of emergency admissions in the outcome year which involve patients identified as at risk of admission. Using the current SPARRA algorithm, the sensitivity observed for those with a risk score of 50+% is 10.5%. This means that approximately 10.5% of emergency admissions within this cohort relate to patients above this risk threshold.It is also possible to calculate a sensitivity measure based on the bed days incurred by those admitted. Bed-day sensitivity at a 50% risk threshold for the current algorithm is 20%.
Q9. How are SPARRA data distributed?
The output from SPARRA is currently made available on a quarterly basis: through SPARRA Online. Quarterly SPARRA scores are routinely uploaded into a SPARRA data mart. A Business Objects universe allows front line users secure and appropriate access to the SPARRA lists for patients within their locality, i.e. GP Practice staff have access to their practice patient data and summary reports.
Note that SPARRA Online provides details of patients with risk scores of 10% or more.
Q10. How frequently are SPARRA scores updated?
Scores are calculated on a quarterly basis and are released at the end of January, April, July and October.
Q11. Who currently receives SPARRA lists?
SPARRA listings are currently made available to a wide range of individuals at NHS Board, CHP and GP Practice level. In some cases, these data are then shared with social work colleagues.
Q12. How can I receive SPARRA data for my area?
For access to SPARRA Online, a SPARRA User Authorisation Form needs to be completed and returned to ISD. Please see the SPARRA Online page for further information.
Q13. Who is the relevant signatory for my SPARRA User Authorisation Form?
- For release to NHS Boards the sponsor should be the Board Caldicott Guardian
- For release to CHPs the sponsor should be the Clinical Director of the CHP
- For release of data to General Practice the sponsor should be a GP Principal in that practice.
- For individuals requiring access to data relating to the registered patients of two or more General Practices, the sponsor should be the Clinical Director of the CHP.
Q14. Who should I contact for further information?
Please contact a member of the SPARRA Team for further details.
