(A-280) Like father, like child: feasibility of linking father and child in CPRD GOLD

Background: There is a growing need to investigate how paternal exposure to medicines prior to conception might affect pregnancy and infant health. Clinical Practice Research Datalink (CPRD) has a mother-baby link to connect maternal and child records in the UK. Although one prior study reported using the CPRD GOLD data source to link paternal and maternal records with child records, a father-baby link is not currently available from CPRD.

Objectives: Investigate the feasibility of linking child and paternal records in CPRD GOLD.

Methods: All newborns and all males aged 16-65 years registered in CPRD GOLD between 01/Jan/2020 and 31/Dec/2024 were identified. We used the general practitioner practice identifier (ID) and the family number (a number assigned to each household by the practice where the patient is registered; family numbers can be repeated across practices).
Newborns were matched to possible fathers with the same practice ID and family number; the potential father had to be registered with the practice at the time of the newborn’s birth. Subsequent additional restrictions were applied: 12 months of paternal registration in the practice before the newborn´s birth, only 1 potential father per newborn, and paternal age between 25-40 years at the newborn’s birth. Proportions of linked newborns were calculated over the study period and annually.

Results: 135,550 newborns and 1,627,789 males aged 16-65 years (potential fathers) were identified. 120,588 newborns (89.0%) had a unique combination of family number and practice ID; the number of newborns sharing family number and practice ID (i.e., seemingly belonging to 1 household) ranged from 1 to 10 (median=1; upper quartile=1). 1,364,216 potential fathers (83.8%) had a unique combination of family number and practice ID; the number of potential fathers sharing family number and practice ID ranged from 1-159 (median=1; upper quartile=2).
49,108 newborns (36.2%) were matched on family number and practice ID with at least 1 potential father registered with the practice at the newborn’s birth. Restricting to pairs with only 1 potential father who was registered at least 12 months before the newborn’s birth date resulted in 31,077 pairs (22.9%). Further restriction to potential fathers aged 25-40 years resulted in 24,545 pairs (18.1%). This linkage proportion was similar across years, from 18.9% of newborns born in 2020 to 17.5% in 2024.

Conclusions: Although simply matching on family number and practice ID resulted in nonspecific matches, by including additional restrictions, possible father-child pairs that could be used in paternal exposure safety studies were identified in CPRD GOLD. Evaluation of the accuracy and representativeness of the linked population is recommended before conducting such studies.