Table 3.

Examples of How False Matches Might Be Produced by Errors in Identifiers or in Data Submitted by Hospitals

Situation	Example
Characteristics of patients
Unconscious patients, frail patients with dementia, patients under the influence of alcohol or drugs, abandoned babies	Unknown dates of birth can result in missing data or default values (see below) (HSCIC 2010)
Unconventional surname	Naming conventions can contribute to linkage errors, because names may be stored differently on different databases, be presented by patients in different ways to different hospitals, and may be misunderstood by frontline staff. In an analysis of 100 records which failed to match, and did not have an NHS number, 37 failed to match because of the name(HSCIC 2009a)
Address out of date	If 40% of patients have not registered with a GP 6 months after moving, the address provided may not match that shown on the PDS (Millett et al. 2005)
“Complex case”	Examples include having no NHS number, invalid PDS record, demographic details out of date, demographic details not supplied by patient, Scottish patient presenting in England for the first time (NHS Connecting for Health 2012)
Misleading information given	A drug user may presents at two hospitals with different names
Visitor	A visitor to the United Kingdom may have no NHS number or postcode
Match by co-incidence	93,000 records shared sex, postcode, and date of birth by coincidence in 2006/2007 (excluding multiple births) (HSCIC 2009b). This could potentially explain records apparently showing two birth episodes separated in time
Multiple births	Multiple births share date of birth and postcode. Prior to 2002, they would have not received individual NHS numbers until they were registered. Prior research has shown that data about the first baby in a multiple delivery are more complete than subsequent babies (Dattani, Datta-Nemdharry, and Macfarlane 2007). For example, an infant death followed by apparent re-admissions might refer to twins
Communal establishments	Shared housing and communal living establishments can lead to false matches, although HES exclude a regularly updated list of communal postcodes from stage 3 of the algorithm. Previously, postcodes creating 10 matches were excluded (HSCIC 2009b)
Errors in data submitted by hospitals
Default dates of birth entered	Default date of birth values (e.g., January 1, 1900) or estimated dates of birth may increase false matching (HSCIC 2010)
Default postcode entered	Default postcodes may be entered, which may or may not follow national guidelines (e.g., recognized default postcodes for homeless people and travelers may be entered as the hospital or embassy postcode by frontline staff)
Multiple births given same identifier	If hospitals give multiple births the same ID number (Dattani, Datta-Nemdharry, and Macfarlane 2007), or leave the field missing, this would increase the chance of false matching
Sharing NHS number	An NHS number can be unverified (not checked against the PDS) or invalid (fail a number check digit calculation) (HSCIC 2012b). Even valid NHS numbers can refer to the wrong patient. A check digit will fail 10% of the time due to typographical errors. Patients can therefore end up sharing NHS numbers (e.g., mother and infant sharing an NHS number might explain infant deliveries coded as births, or apparent simultaneous admissions to different hospitals on the same day)