
Articles
176
www.thelancet.com/psychiatry Vol 11 March 2024
Methods
Study design and participants
This cohort study used three consecutive versions of the
national dataset of secondary mental health care provided
by NHS England, which together included mental health-
care episodes from April 1, 2006, to March 31, 2019, linked
at patient level to Hospital Episode Statistics, which is the
administrative database of all care episodes in general
NHS hospitals,9,10 and birth notifications from the
Personal Demographic Service, which is the national
master database of all NHS patients in England
containing patient details such as name, address, date of
birth, and NHS number.11 Records of mental health-care
episodes between Dec 1, 2015, and March 31, 2016, were
not available for technical reasons.
Records from the Hospital Episode Statistics database
included patient demographics, admission dates, diag-
noses coded according to ICD-10, and procedures coded
according to OPCS Classification of Inter ventions and
Procedures version 4 (OPCS-4) codes (appendix pp 2, 4).12
Each record of a maternity episode in the Hospital
Episode Statistics database contains a set of additional
data fields, often referred to as the maternity tail, which
contain specific information, including birthweight,
gestational age, and mode of delivery.9 Birth records from
the Personal Demographic Service contained additional
information on stillbirth, gestational age, sex of the baby,
and birthweight, which was used if Hospital Episode
Statistics data were missing. Ethics approval was provided
by the NHS Health Research Authority (19/SW/0218) and
data were provided by the NHS Data Access Request
Service (DARS-NIC-376141-W5D3L-v0.9).
For this study, from Hospital Episode Statistics records
and Personal Demographic Service birth notifications, we
identified maternity episodes of all women eligible for
inclusion (aged ≥18 years) with a recorded gestation
(≥24 completed weeks) with an onset of pregnancy from
April 1, 2016, and a singleton birth up to March 31, 2018.
This inclusion period was chosen because it allowed us to
determine the mental health outcomes according to the
most recent version of the secondary mental health
services dataset for all included women (appendix p 2).
Women who had multiple births were excluded for
practical reasons; first, their number was small, and
second, including them would have made the statistical
analysis more complex because it cannot be assumed
that the outcomes of babies of a multiple birth are
independent observations. If there were multiple
maternity episodes for the same woman during the study
period, one maternity episode was selected using a
random number generator to be included in the study to
minimise selection bias and, again, to avoid the need to
use more complex statistical methods appropriate for the
analysis of non-independent outcomes.
To determine the onset of pregnancy, we subtracted the
gestational age at birth minus 2 weeks (or 38 weeks if
gestational age was not available) from the date of birth.
Women were included if they had a pre-existing mental
illness, defined as a contact with any form of secondary
mental health care, within a look-back period of 10 years
before the onset of pregnancy.8 This 10-year look-back
period was chosen on pragmatic grounds, because the
historical mental health services data were available from
April 1, 2006.
Procedures
We distinguished three levels of secondary mental health
care as a proxy for the severity of the mental disorder.
The first level (hospital admission) was an admission to a
psychiatric ward, including generic psychiatric wards,
mother–baby units, or secure wards. The second level
(crisis resolution team) was the involvement of a mental
health crisis resolution team providing intensive
treatment at home for an acute mental health crisis
otherwise needing hospital admission. The third level
(community care) was any other secondary mental
health-care contact, including day care and outpatient or
community-based care. More information can be found
in a previous paper.3
We did not use recorded diagnostic information to
distinguish the nature or severity of the women’s mental
illness because of the high rate of missingness of
diagnostic information, especially in records of the crisis
resolution team or community care. It has been widely
recognised that mental health diagnoses are often not or
incompletely recorded in administrative datasets of
mental health care.13
Information on the regional availability of community
perinatal mental health teams was collected at the level of
the Clinical Commissioning Groups.14 These groups
were NHS bodies responsible for the planning and
commissioning of health-care services for their local
area between April 1, 2013, and July 31, 2022. Initially,
211 Clinical Commissioning Groups were established
in 2013; this number decreased to 135 in 2020 because of
mergers between groups. In this study, we define the
regions according to the 207 Clinical Commissioning
Groups present in July, 2017, given that these reflect as
accurately as possible the situation for the women
included in our study. Typically, these regions cover areas
with a population of 250 000 people. We surveyed the
Clinical Commissioning Groups to determine the date of
community perinatal mental health team implementation
for each region, defined as the earliest date that at least a
dedicated psychiatrist, psychologist, and specialist nurse
were in place. We considered a community perinatal
mental health team to be available to women if the date
of implementation in their Clinical Commissioning
Group region was before the onset of pregnancy.
Outcomes
The primary mental health outcome was acute relapse
(psychiatric hospital admission or crisis resolution
team contact) in the postnatal period (the first year after
See Online for appendix