The GPRD was initiated in 1987 and is the world’s largest anonymized, longitudinal patient electronic medical records database providing clinical information based on GP records. The GPRD data contain approximately 46 million patient years of follow-up representing 10.11 million unique patients.184Over 460 general practices in the UK are currently submitting data to the GPRD on 3.23 million patients or approximately five percent of the UK population.154, 155, 184 The patient population is representative of the regional, age and gender distribution of the UK population.184
Members of the UK’s National Health Service (NHS) act as the main means of access to and record holder for all forms of health care provision within the NHS. Practitioners in the GPRD tend to be part of larger rather than smaller practices. Practitioners enrolled in the GPRD must follow a recording protocol ensuring that significant clinical contacts are entered into the computer record. These contacts include all events resulting in hospitalization or referral to any specialist. The outcome of the referral is also recorded. Any significant test results are recorded in the GPRD. All events resulting in a prescription or withdrawal of treatment are recorded. Any events that the patient will consult with the practitioner on more than one occasion (childhood diseases, pregnancy) are often recorded multiple times by the practitioner.185
The MHRA has put in place specific recommendations regarding recording of
of the pregnancy when known. This includes positive pregnancy test results and any referral for ante-natal care. Additional information concerning significant abnormalities or
complications of the mother or her fetus detected during pregnancy are also recorded. The outcome of the pregnancy, including the date of delivery, any congenital malformations of the baby, and where relevant, a record of neonatal death, are recorded.185 Free text may be recorded by GPs to further detail the patient’s medical conditions. Diagnoses are recorded using Read Codes (1996-current) and a modified version of the Oxford Medical Information System (OXMIS: 1987-1999). For this project data were selected from calendar years 1987 through September 2004, and thus require both the OXMIS and Read Code systems.
GPs code clinical information for their practices using the greater than 80,000 Read Codes which cover a wide range of topics in categories such as signs and symptoms, treatments and therapies, investigations, occupations, diagnoses, drugs and appliances. However, the GPRD also uses the 18,000 codes from the Oxford Medical Information System (OXMIS). Because these two coding systems co-exist within the GPRD, a cross- classification variable was developed called the GPRD Medical Code that allows
investigators to query the system using Read, OXMIS or these GPRD Medical codes, with the latter being the most efficient for research purposes. The GPRD Medical codes allow for consistency over time in defining disease outcomes and thus they were used to conduct all analyses for this dissertation.
We created a data file containing all electronic patient records with a NTD code within the GPRD between January 1987 and September 2004. All the clinical, referral, diagnostic and screening test (and results), immunization and therapy events for this cohort were downloaded. The data set was created using the GPRD Business Objects data acquisition
system at the RTI-Health Solutions London, UK office and is in the form of tab delimited text files that were then converted into SAS datasets. The SAS datasets were used for all analyses. The medical code listings we used to identify NTD patients are listed in Appendix B. As many NTDs are associated with a variety of syndromes and often known by a variety of names, GPRD Medical codes were searched using a list of key words developed through an extensive review of the literature in addition to work by Jones40 and Moore.81 This list of codes is meant to be inclusive of all codes for anencephaly, craniorachischisis,
encephalocele, encephalomyelocele, meningocele and spina bifida.
Pregnancy codes were identified through a key word search of the GPRD medical code dictionary. Codes were divided into two categories: End-of-pregnancy (EOP) events and pregnancy-care-markers (PCMs). EOP events are those events that represent the final outcome of a pregnancy, such as live births, stillbirths, miscarriages, spontaneous abortions, elective terminations, multi-fetus deliveries and pre- or post-term deliveries. PCMs include any event that describes the delivery of care relating to pregnancy prior to an EOP event. Examples include positive pregnancy tests, alpha-fetoprotein tests, obstetric ultrasounds, amniocenteses, visits related to pregnancy, pregnancy complications, threatened abortions, abortion referrals or counseling, and obstetric hospitalizations.
In addition to the keyword search of the code dictionaries, we created longitudinal patient histories for a subset of 10,000 women with at least one EOP code and visually reviewed them for previously undiscovered pregnancy related codes. These methods resulted in a total of 5,266 codes that were potentially associated with pregnancies (Appendix C). Our final list of codes consisted of two subsets of codes: one containing 1,691 PCMs codes (Appendix D),
and another containing 1,059 EOP codes (Appendix E). We excluded the remaining 2,516 codes as they represented post-natal care, or were deemed non-specific.
C. Procedure For Pregnancy Identification