The Master Person Service (MPS) helps us increase the amount of usable, better-quality data available to support research and planning. Patients can visit multiple places where they register to receive care or treatment. This information is stored and recorded in various systems around the country. MPS aims to match the right person with the right record.
MPS takes the demographic information contained in a person’s health and care records and matches it to their unique NHS number to confirm their identity. A bespoke deterministic linking approach is used to provide a single best result for each record against the Personal Demographics Service (PDS) or against the MPS record bucket that contains previously unmatched records.
The matching process is comprised of several steps of increasing computational intensity, designed to address the challenges provided by data sets of different degrees of quality.
As a result of the matching process, a person identifier is assigned to the record. This might be an NHS number if the record was matched in PDS, or an alternative identifier called MPS_ID if the match was found on the MPS record bucket or there was enough information to create a new one. Records that could not be matched against either databases are assigned a randomly generated one-time-use ID.
Like any data linkage method, MPS cannot provide perfect matching. There are risks of both failing to match a record (false negative) and matching to a record incorrectly (false positive). The performance of MPS is determined by both the algorithm itself and the quality of incoming data.
MPS operates in the same way for all data sets and is not tuned to any particular use case. For example, where records reliably have accurate NHS numbers attached, MPS will provide a correct match with high confidence. Where solely relying on other personal identifiers, such as name, postcode, gender or date of birth which may be incomplete, inconsistently recorded or duplicated across the population, the algorithm will be less able to return a correct match in all cases.
Mature health datasets, where identity is typically validated in a healthcare setting at point of recording (such as HES), have higher levels of matching accuracy through MPS for most records. Performance for other datasets may be variable.