By Dr Tom Foley.
Dr Gerry Morrow is Medical Director of Clarity Informatics, who aim to improve patient care and outcomes using data and analytics. Clarity offer a Quality Improvement Service (QIS) in 4 regions of England, that covers 11 clinical focus areas including Sepsis, acute myocardial infarction, coronary artery bypass grafting, dementia, first episode psychosis, pneumonia, heart failure, stroke, hip and knee surgery, acute kidney injury and chronic obstructive pulmonary disease.
Data used by Clarity is collected from both primary and secondary care using various sources.
Secondary care data sources often depend on local agreements on data sharing. For secondary services this is usually Secondary Uses Service (SUS) or Hospital Episode Statistics (HES) data and comes from the Health & Social Care Information Centre (HSCIC), trusts, intermediaries, Commissioning Support Units (CSUs) and health information exchanges. Contained within this data is a wide range of pseudo-anonymised information concerning each patient’s episode, including demographic information, site codes, hospitals involved, type of interaction (inpatient, outpatient or day case), ICD10 code (International Statistical Classification of Diseases an Related Health Problems) or OPCS code (Office of Population Censuses and Surveys Classification of Surgical Operations and Procedures. It also includes outcome data regarding readmissions, length of stay, discharge destination including whether discharged alive or dead. This type of data tends to be 6 weeks retrospective and therefore slightly out of data.
Primary care data is collected directly from the GP practice. In this setting, there is a “cradle to grave” patient record. These are based on READ codes. GP data is queried using SQL (Structured Query Language) and the results are aggregated at GP level before being extracted. GPs provide consent for the duration of the project and do not need to approve individual queries. This process does not interfere with the day to day running of the practice. Queries are made on a monthly cycle, and the data can be reviewed over time to identify groups/cohorts of patients for each illness, for whom the quality of care has or has not improved.
Secondary care data is augmented with clinical data relating to each patient episode. This is added by human coders, who review the patient notes, which are often in paper form. This adds information and accuracy. For example, this makes it possible to review how quickly someone was admitted to a particular ward, how timely their X-ray, CT scan or other investigation was undertaken, how quickly they received antibiotics, whether they had physiological scoring e.g. CURB score, etc.
Those process measures can be provided as a series of reports to participating trusts, clinicians, managers and Clinical Commissioning Groups (CCGs). Outcome and process measures can be linked to provide useful management information, so for example, it is possible to show that timely administration of the first dose of antibiotics improves outcomes.
Data is validated to ensure that there are no duplications or gaps. There are checks with ICD10 or OPCS codes to ensure that the patient is appropriate to that cohort, and if this is not the case, that data is not stored.
Clarity adheres to all the relevant IG standards. Data is transferred by secure file transfer protocol (FTP) or via N3 (the national broadband network for the NHS). Data is securely held and partitioned so that each area can only see their own data.
The Care.data project has presented new information governance challenges. GPs and patients have become more concerned about confidentiality and data access issues. The only way to gain the trust of the public back to this initiative might be through an opt-in approach rather than the planned opt-out. A centralised repository of all the information proposed within Care.data would be a fantastic resource and would allow better benchmarking of services.
When a patient signs up to a GP practice, there is a process to explain that their data may be used for audit purposes. Most patients already assume there is sharing of data for this purpose. Patients can and do decline, just as some decline to be uploaded to the national spine. No patient identifiable information is transferred, it is all pseudo-anonymised, and Clarity receive a randomly generated number and do not know who it bellows to. The GP can have a look up table so they can look up an individual patient but Clarity do not have access to that information and do not receive an NHS number, date of birth or address.
Linking secondary and primary care data is difficult as pseudo-anonymisation gives different patient identifiable numbers for each. Therefore tracking a patient across this boundary can be challenging but it can be done. It is made easier by automating the process where possible.
National figures used to benchmark tend to be high level such as trends in mortality. For example, it is possible to give disease specific comparisons for mortality levels in the North East compared to the national level.
Augmented data gives additional granularity. An example of this is with Cardiothoracic surgery where reports could include not just mortality but also number of blood units transfused, dehiscence of wounds or significant complication, long stays, multiple visits to outpatients, etc.
Data controllers can be based at a provider or can be an intermediary such as the academic health science networks which often correspond with geographical regions. These bodies are usually in control of the data and the direction of the projects
Comparative Effectiveness Research
This work can be used for comparative effectiveness research, for example it would be possible to compare anticoagulation therapies given for VTE prophylaxis. The HES and SUS data are augmented to give the information required to make this possible. Drugs lists are updated on a monthly basis so that new therapies are incorporated into the data. Augmentation is completed at the provider level, usually within the audit department or by a data entry clerk. This process has been adapted to make data collection as simple as possible, with pre-populated answers and drop down menus and this has reduced the coding time to around 5 minutes per patient.
With this information it would be possible to track outcomes such as readmission for specific treatments and evaluate drug A vs drug B. In many cases there would be small numbers in individual providers but it would be possible to aggregate this across regions. There may be some difficulties due to regional variation in practice and each study would need to be aware of the clinical priorities in different regions.
Reports can be used to feedback where there may be gaps in care. Reports are often used in multidisciplinary team meetings involving clinicians, coders, nurses and managers to feedback how well they are performing and can compare against other areas. Usually there is a spread of standards with some outliers who are either doing fantastically well or in some cases badly. This method provides an opportunity for the sharing of best practices and drives improvement. These reports can be very powerful tools.
It is possible for providers to generate their own custom reports using this data and Clarity can provide the tools to help them to do this. There are options for pre-canned reports or they can generate more detailed reports. More detailed reports can become complex and often it is data analysts who would undertake this work. They can use the data in a number of ways to produce high quality information that can be used to improve care. It is possible to use logical operators and filters to generate insights into particular groups, for example, those aged over 60, from a specific geographic location, admitted in a certain time frame with a particular condition.
These reports can be used by providers themselves but also can be of interest to organisations such as the Care Quality Commission and Monitor who are interested in finding outliers. There are cases where mortality may be high in a specific region but the data can show that care provision is actually appropriate and that standards have been adhered to. This can be used to demonstrate that this higher mortality rate may be due to population variation, for example, in areas with an elderly population.
There are not many services that can provide this level of granularity. Studies can be demonstrated to be of statistical significance, and this would be done at the analyst level. However, if examining very specific areas there may not be sufficient numbers to reach statistical significance and it is often useful to aggregate a region to obtain this. In circumstances where this is not possible the raw figures can still be of use.
There have been notable statistically significant improvements in the areas where we have been working, for example, pneumonia related mortality in the North West has reduced by 2%, sepsis has reduced 4% and readmissions have reduced in heat failure. These benefits may be due to more than the sum the study parts. There can be additional effects that are unintended and there is a lot more going on than just collecting and reflecting the data.
Data collected from SUS and HES has already been standardised to their particular forms and therefore there is no problem with its interoperability. For primary care data which comes from a variety of systems, such as EMIS or SystmOne, their underlying data models and architecture are the same as they are based on READ codes however there was a need to develop a specific system to interact with each individually.
In primary care, finding the right incentives can be more challenging. This is because GPs are already working at capacity and have various priorities related to the Quality Outcomes Framework and other incentives. Giving them more targets would not be popular, if the demand is driven by practices for particular conditions that they want to explore then it is more likely to be successful.
In secondary care, nationally mandated projects would be very helpful, for example a project exploring sepsis across all of NHS England. Clarity would be very interested to be involved with projects like that.