Recommendations that could improve Data Quality in Primary Care
- Pick lists of codes
- Able to update these lists at a national scale. Allows for national control over the default preferred search terms; allowing these to be standardised and related to SPIRE
- Different list of codes can be presented depending on healthcare provider’s current role.
- Lists of preferred (recommended / ‘formulary’) codes are shown first (and clearly displayed or highlighted as recommended codes) but it is possible to search from larger number of codes if relevant code is not found.
2. SNOMED CT search terminology integrated within the Electronic Health Record
The ability to quickly search SNOMED CT within the electronic health record is fundamental for this to be used.
3. Natural Linguistic Programming
Medical records tend to be free text with some coded data. Natural language processing functionality allows use of computer to process notes and code them. If this is performed across the nation this has the ability to improve consistency of coding.
4. Search functions are user friendly and fit for purpose
Functional searches for codes that allows for searching for codes with an autocomplete (suggested word completion abilities). It is imperative that this is functional, user friendly and fit for purpose. This will assist in performing accurate and consistent coding.
5. Artificial Intelligence
Artificial Intelligence (AI) has the potential to improve healthcare and standardise notes. For example initial contacts to general practices could be dealt with via an AI chatbot; at the end of this encounter a checking method is performed which asks if the summary is correct – the coding of this encounter would be standardized. In addition AI has the potential to assist with retrospective notes summarizing.
6. Adhere to interoperability standards
Health and social care data currently exists in silos. Adhering to best practice interoperability standards (such as SMART on FHIR), allows data to be shared between various settings and more comprehensive data to be provided.
7. Automated coding of data received from external sources
The ability to automatically code data that is received from external sources such as immediate discharge letters, secondary care correspondence, immunizations and screening tests would be valued.
8. Recording of consultations via video with machine learning interpreted coding
Ability to record both visual and audio of consultations, whether remote or in person and allow a machine to code accordingly. This allows playback of the consultation by patients (helping them improve health literacy). Utilising machine learning will allow these also to be coded consistently.
9. Integration of decision support tools
Decision support systems have the potential to improve quality of care and to reduce unnecessary variations in clinical practice. These systems rely in part upon adequately coded diagnostic data so they function correctly. By providing clinicians with the ability to utilise clinical decision support tools and making them aware that good quality data is important for these provides the ability to improve data quality.
10. Voice Record keeping
Current software relies upon a keyboard and a mouse largely to input data. Having the ability to ‘dictate’ / speak to the machine and add relevant information readily in the correct place.
11. Assist patient orientated accessible services
Allowing patients to interact with the electronic health record will allow them to show greater interest and potentially improve the quality of data. To do this having functionality so that the end user can upload relevant healthcare data (such as from wearable technologies), rather than just view a high level summary would allow a more comprehensive record to be obtained.
12. User centred design for healthcare provision, planning and targeted notes review
System that reinforces good practice to maintain disease registers, with user centered design ensures this is fit for purpose. For example identifying patients who may have missing disease data suggested from prescribing data (such as a missing diagnosis of diabetes if receiving insulin).
13. Functional audit tools with information visualisation
Readily able to configure or link to relevant dashboard that displays metrics of data quality. In addition should include the ability to analyse metadata to assist data cleaning.
14. Promote unified user (healthcare provider) experience
Utilising cloud storage users settings can be configured in a way in which the user most prefers. In addition this means that when working across multiple sites and different health boards the user has the ability to have their system set up in a manner they prefer.
15. Coding of encounter type
Type of encounter (such as home visit, telephone call, administration) can be set by practice as a default (relating to where this sits on the appointment system). Improvements to the way the type of encounters are recorded are required as currently clinician’s tend not to change this.
16. System navigation and usability are optimised so coding requires minimal time
Clinicians and administrative staff are short of time. When coding a disease it is important that this can be performed readily with the least number of mouse clicks / menus to select from. By making it easier to record data this could improve data quality.
17. In-built standardised setting of code priorities
Standardised setting of code priorities (and possibly sensitivity) with the ability to define what appears in a summaries / referrals for different clinics / specialties could be useful. For example the ability to include obstetric delivery codes for referrals to gynaecology but exclude these to other specialities referrals where not relevant.