Data resource profile: Adult Psychiatric Morbidity Survey (APMS)

Abstract

The Adult Psychiatric Morbidity Survey (APMS) monitors the prevalence of mental illness and access to treatment in the general population for England’s National Statistics. The primary aims of the survey series are: To estimate the prevalence of different mental disorders in England’s general population. To establish what proportion of people with a disorder are in receipt of treatment. To produce temporal trends in prevalence and use of treatment. To compare circumstances of people with and without a mental disorder. The series consists of four repeat cross-sectional surveys. The first two covered Britain (England, Scotland and Wales) and were conducted in 19931 and 20002 by the Office for National Statistics (ONS). The 20073 and 20144 surveys covered England only and were conducted by the National Centre for Social Research. The later surveys had no upper age limit (this was 64 years in 1993 and 74 years in 2000). A 2021 survey is planned. Each survey consisted of a first and second phase. The second phase was co-ordinated by the University of Leicester and carried out with a subgroup of phase-one participants. England’s Department of Health and Social Care has been the main funder throughout the survey series, with the most recent surveys commissioned via NHS Digital. Each survey underwent the relevant ethical review, most recently with the West London National Research Ethics Committee 14/LO/0411. The general population surveys are part of a wider series of studies examining the mental health of specific populations. These have included people living in institutions,5 the homeless population,6 prison population,7 carers8 and minority ethnic groups.9 There has also been a series of mental health surveys of children and young people in England,10,11 and these are described in a separate Data Resource Profile.12 Each wave of the APMS used a similar stratified random probability sampling design to produce a sample representative of the population living in private households. The first stage of sampling involved selection of addresses from the Postcode Address File, which covers around 97% of households in England.13 People living in temporary housing, institutional or communal establishments or sleeping rough, were not sampled. Although mental illness rates may be elevated in these populations,14 they comprise <2% of the population and their exclusion should not impact on overall estimates.15 After the mailing of invitation letters with information about the survey, trained interviewers visited each address. They established whether households were private households. One resident (aged ≥16 years) was randomly selected in each eligible household. Fieldwork took place from April to November 1993, March to September 2000, October 2006 to December 2007, and May 2014 to September 2015. Response rates are given in Table 1. In 1993, 10 108 people aged 16–64 years, and in 2000, 8886 people aged 16–74 years, were interviewed in England, Scotland and Wales. In 2007 there were 7461 participants aged ≥16 years, and in 2014, 7546, in England only. Weights took account of selection probabilities (such as the under-sampling of those living in multi-person households) and non-response, to render results representative of the household population aged ≥16 years at the time of each survey. ONS mid-year population estimates were used for population control totals for age-by-sex and region. The full weighting strategy is described in the survey reports. APMS response rate APMS response rate The process for selecting which participants should be invited for a phase-two assessment varied between survey waves. In 1993 and 2000, all those screening positive for psychosis or a personality disorder were followed-up for phase two. In 2007 and 2014 more granular sampling fractions were applied. Further refinements to the sampling fractions were applied in the 2014 survey, such as varying the sampling fractions applied for men and women. In the 2014 survey, the probability that a phase-one participant would be selected for a phase-two assessment was the highest of two calculated probabilities: psychosis probability and autism spectrum disorder (ASD) probability. These disorder-specific probabilities were corrected for using disorder-specific weights. Phase-one interviews were conducted by trained research interviewers in people’s own homes and averaged 1.5 h in duration. The phase-two interviews were also carried out in participants’ homes and took a similar amount of time. A small token of appreciation was given to participants in each phase. Since 2000, interviews have been conducted using computer-assisted personal interview, whereas in 1993 data collection was by pen and paper. Most of the questionnaire was administered face-to-face, with some sensitive information collected through self-completion. A variety of types of mental disorder were either assessed or screened for. These included relatively common conditions such as depr