Aim

Population-based observational studies using routinely collected healthcare data allow analysis of variation in service use and outcome to be studied. This data linkage study aims to examine the impact of the pandemic on patterns of NHS services utilisation for people from the most deprived communities in North East and North Cumbria, England, and determine whether these patterns were associated with health-related outcomes.

Objectives

The specific objectives of this study are:

1. To investigate NHS service utilisation (i.e., settings and pathways of care) where people were in contact with NHS-funded secondary mental health services in the year prior to the lockdown across the deprived population and in specific groups such as the elderly or ethnic minorities.
2. To quantify changes in their mental health service utilisation between multiple time periods across pre-, during- and post-lockdown including out of lockdown in England.
3. To identify the specific factors at the individual and general practitioner (GP) -levels that contribute to differences in the use of mental health services.
4. To quantify patients’ health outcomes among the cohort between the multiple time periods.
5. To estimate the associations between mental health service utilisation and patients’ health outcomes
6. To explore the contributions of individual- and GP practical-level factors to associations between mental health service utilisation and health outcomes, establishing when, where and for whom mental health services may be effective.
7. If data is available, to conduct comparative analyses using data from the less deprived communities in North East and North Cumbria, England.

Design

This retrospective study will link seven routinely collected healthcare data sets in England between March 2019 and March 2022, to assemble a retrospective observational cohort of adults (over 18 years) in contact with mental health services from the most deprived communities.

Ethics statement

Ethical approval was granted by Health Research Authority Research Ethics Committee (REC) in October 2022 (22/NS/0080). Under Regulation 5 of the Health Service (Control of Patient Information) Regulation 2002 (‘section 251 support”), this study has been approved to process confidential patient information without consent by Confidentiality Advisory Group (22/CAG/0093). The Confidentiality Advisory Group is an independent group of lay people and professionals which provides expert advice on the use of confidential patient information without consent in the United Kingdom. Data sharing agreements will be in place. Applications for NHS routine data will be made to NHS Digital via Data Access Request Service (DARS). To protect privacy and confidentiality, approval for the linkage of health data is provided under strict conditions for the storage, retention and use of the data. Only anonymised data will be shared with the approved research team for analysis in this study. CAG and REC approval letters are provided (see online supplementary appendix).

Setting

The North East and North Cumbria (NENC) is the setting for this research, which is consistently ranked as having the highest poverty levels and the lowest health outcomes of any region in England. Participants in this research will be within the 20 GP practices identified as “Deep End” that fell into the 10% most deprived GP practices in England against the Index of Multiple Deprivation (IMD, 2019), prioritising geographical representation from Deep End GP practices across the NENC region. These GP practices have between 57.6% and 95.9% of registered patients living in the most deprived 15% of IMD Lower Layer Super Output Areas (LSOAs). The findings of this research could understand the effects of COVID-19 on deprived population and guide the development of the region’s Deep End Network.

Base cohort

The study will include adults over 18 years because there are significant differences between the organisation of child and adolescent mental health services and adult mental health services in England. The base cohort from the Deep End GPs will consist of patients who are currently registered with the practices and who were referred or self-referred to NHS-funded secondary mental health services or Improving Access to Psychological Therapies (IAPT) services between 23 March 2019 and 22 March 2020. Data for mental health referrals is defined as instance where a patient’s care was directed to mental health services, including referrals to secondary care and community care. A flow chart of how the base cohort will be formed and the administrative data sets to be linked is presented in Fig 1. Inclusion in the cohort may be modified depending on the specific research question being addressed. Patients who had mental health conditions without having been referred or self-referred before 23^rd March 2020 or who have moved out of the NENC region will be excluded in this study. For those who have moved into the NENC region during the pandemic, we will examine the impact of population movement on the outcomes of interest and control for this factor in the statistical analysis in order to isolate its effect on the results. The data linkage is expected to be completed by February 2023 but is subject to timely approvals and linkage from external agencies.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Formation of the linked datasets.

https://doi.org/10.1371/journal.pone.0283986.g001

Sampling frame

We will use a total enumeration approach to include all participants fitting the inclusion criteria in each of the identified GP practices. Of the 34 local Deep End GP practices, four were in significantly rural areas, two were in urban with significantly rural areas, and 28 were in predominantly urban areas. To maximise the rural sample size, all of the six GP practices in either significantly rural or urban with significant rural areas were selected. 14 GP practices with the highest IMD scores from the 28 GP practices in predominantly urban areas were selected.

The Mental Health Statistics (England, 2021) indicated an estimated 5% of the national population had contact with NHS-funded secondary mental health services during 2020/21. From the Fingertips data (2019/20) [15], we expect 107,808 adults registered across the 20 GP practices. Therefore, we estimated a total sample of 5,390 adults in contact with NHS-funded secondary mental health services in the 20 GP practices. As a general assessment of power to detect changes in waiting time comparing before and after the COVID disruption to services, we can report power for the simplistic situation assuming a paired t-test was to be used to detect changes in waiting time. 5,390 participants would give us 95% power at the two-sided 5% significance level to detect a true difference of 0.05 standard deviations in waiting time. This sample size calculation is a guide, as in reality the data will consist of repeated measurements of the participants and more appropriate methods will be applied for the analysis which take account of the repeated measurements and nested nature of the data.

Time periods of the COVID-19 lockdown in England

Time periods [16] studied are

• Pre-lockdown (23^rd March 2019 – 22^nd March 2020)
• First national lockdown (23^rd March– 3^rd July 2020)
• Minimal lockdown restrictions (4^th July– 13^th September 2020)
• Reimposing restrictions (14^th September– 4^th November 2020)
• Second and third lockdowns Lockdowns 2 and 3 were combined because of a short interval between them (5^th November– 7^th March 2021)
• Leaving lockdown (8^th March 2021–30 June 2021)
• Out of lockdown (1^st July 2021 – 22^nd March 2022)

Routine data sources and linkage

On identifying the base cohort, North of England Commissioning Support Unit (NECS) will 1) extract primary care data from the identified GP practices; 2) send a list of identified patients with their personal information (i.e., NHS number, name, and date of birth) to request secondary and community care data from NHS Digital; 3) perform data linkage using the personal information for each patient which is a required field in all datasets; 4) create unique identifiers for all participants and will apply a pseudonymised code to the datasets; and 5) supply a pseudonymised linked data to the researchers for analysis. The linked data will be individual level data available from 23^rd March 2019 to 22^nd March 2022. NECS will retain the key for the pseudonymisation, which will not be released to the research team.

Aggregate data

GP practice-level data will be extracted from Fingertips [15], which is a publicly accessible web tool (https://fingertips.phe.org.uk/) maintained by Public Health England. It provides access to a wide range of local public health data presented as thematic profile. Practice-level factors will include the GP practice profiles, mental health profiles, and overall achievement.

Measurements of outcomes

Changes to mental health services utilisation in response to COVID-19 and their impact on health outcomes will be identified from a rapid review of the literature [24]. In addition, the study may investigate selected outcomes for which there is unknown evidence of an association with service utilisation.

Mental health service measures will include:

• Referrals to new services (e.g., type of referrals, waiting times, progression from low to high intensity treatment)
• Waiting time is defined as number of days that a patient waited in a referral to treatment period.
• Access to mental health services
• We define “access to mental health services” as having at least one recorded contact (whether attended or not) for a mental health reason.
• Modes of contact with services
• Pattern of mental health services utilisation

Defined as duration of service use (e.g., the number of GP consultations, the number of admitted (inpatient) days, the number of days with contact with a healthcare professional)

• Treatment adherence/completion (i.e., the number of attended, cancelled and non-attended appointments)

Non-attendance is considered as an inefficient use of health service resources

• Prescription of antidepressants

Health outcome measures related to mental health services will include:

• Mental health (e.g., depression, PHQ-9; anxiety, GAD-7; HoNOS)

Depression symptom severity was measured using the Patient Health Questionnaire 9-item version (PHQ-9) [25], where scores of 10 or above indicate caseness for depression, and a reduction of 6 or more points on the scale indicates reliable improvement in depression symptoms.

The Generalized Anxiety Disorder Scale 7-item (GAD-7) [26] is the main measure of anxiety symptoms used in IAPT services. Caseness is defined as scores of 8 or above, and a reduction of 4 or more points indicates reliable improvement.

The Health of the Nation Outcome Scales (HoNOS) is a clinician rated instrument comprising 12 simple scales measuring behaviour, impairment, symptoms and social functioning for those in the 18–64 years old age group.

• Physical health (e.g., heart disease, diabetes, or diabetes)
• Self-harming
• Psychotropic medication (outcome or service/treatment use?)
• Common mental disorder (CMD) symptoms
• Mortality: premature mortality, time to death
• Time to first readmission
• Use of A&E, A&E readmissions within 30 days
• Morbidity measured as unplanned assessment by emergency care and inpatient admission

Measurements of confounders/mediators

Individual-level clinical characteristics will include:

• Diagnoses
• Prescription
• Hypertension
• High blood cholesterol
• Long-term conditions/co-morbidities

Individual-level characteristics will include:

• Socio-demographics (e.g., age, gender, ethnicity, etc.)
• Smoking status
• Alcohol intake
• BMI
• Physical activity

GP practice-level characteristics will include:

• List size
• percent of satisfaction
• percent of a long-standing health condition
• percent of caring responsibility
• percent of being in paid work or in full-time education
• percent of being unemployed

Planned statistical analyses

In all analyses, multiple confounding variables will be controlled for as appropriate. The below analyses will be undertaken for the total cohort and subgroups (e.g., age, gender, ethnicity, or socioeconomic backgrounds). All analyses will be carried out using Stata version 17.0 [27].

Descriptive analysis will be used to examine socio-demographic characteristics (e.g., age, sex, etc) and clinical characteristics (e.g., diagnosis, prescriptions, etc) for the cohort within the 12-month period prior to the first lockdown. Comparisons will be made using the t-test, rank-sum and chi-square where appropriate. A p<0.05 is considered as significant. If significant health service utilisation imbalance is detected on a particular variable at patient level, and that variable is correlated with outcomes at a level of 0.3 or higher, the variable will be included as a covariate in regression analyses.

Multi-level regression analysis will be used to quantify mental health service utilisation among the cohort at the different time periods of the lockdown and to assess individual- and practice-level characteristics as predictors of health service utilisation because of the hierarchical nature of the data. A three-level model will be constructed (patients clustered within practice, clustered within time period). The model makes two assumptions: 1) that lockdown rules are similar across different geographic areas, and 2) that there are random effects across practices, which allows each practice to have its own dynamics while still taking all information into account when estimating model parameters.

In this study, latent growth models will be applied within a structural equation modelling framework to analyse the relationship between changes in health service utilization and health outcomes. The growth trajectories will be divided into segments representing each time period, and separate growth curves will be fitted to each segment simultaneously. Additionally, the change process of two or more variables will be examined to determine how they are related to one another (e.g., the change in waiting time and the change in health outcomes). Individual and practice-level characteristics will also be included as predictors of health outcomes.

Comparative analyses

This study focuses on people living in the most deprived areas, and we intend to compare their situation to that of people living in the least deprived areas if we can access data on people living in the least deprived areas. This part of the project is at earlier stage of development and the detailed research is still subject to funding and approval. Comparative analysis offers the prospect of distinguishing the contributions of individual factors to geographic variation in health outcomes.

Missing data

For primary analyses, missing data will not be imputed if the values are missing completely at random (MCAR). If it is concluded that data are not MCAR, multiple imputation using chained equations will be performed by creating 10 impute data sets under the assumption of that the missing data are missing at random (MAR). We will use the stratification-variable as well as other known predictive outcomes in the multiple imputation to estimate the missing values. The analysis will be presented as a pooled summary of the results from the 10 datasets. It is not uncommon that missing data and missing variables are discovered only after initiation of research using electronic health record data, making it possible for us to deviate from our original research plan. If so, details of deviation from the protocol will be reported and reasons for the deviation and the implications on the research and conclusions will be discussed. Sensitivity analyses may be reported to evaluate the potential impact of missingness of data and representativeness of the study population.

Main findings

This study aims to examine the impact of the COVID-19 pandemic on patterns of NHS services utilization in the most deprived communities in England. It will determine whether these patterns are associated with health-related outcomes and wellbeing. The findings of this research could provide valuable insights into how the pandemic has affected the utilisation of NHS services in disadvantaged communities and could inform policy and practice in the field of public health. The study could also help to improve our understanding of individual and situational factors that predict mental health-related outcomes and wellbeing, potentially leading to more personalized interventions in practice. Additionally, the patterns of healthcare utilisation identified in this study could improve our knowledge of patient needs and enhance healthcare delivery for disadvantaged communities. This research could serve as a starting point for further studies on this topic, potentially leading to more effective interventions and policies in the future.

Strengths and limitations of the study

With data spanning multiple years and the ability to link records across different services, this study makes it possible to explore the healthcare utilisation history of patients across multiple healthcare sectors both before and after the lockdown. This study can identify the types of referrals to healthcare services made at different points in time facilitating assessment of health service usage and recommendations for improving patient care pathways, and thus maximise clinical efficiency and efficacy in the use of resources. The study findings will be assessed and presented to direct policy relevance, and we plan to share them directly with those bodies working in these areas at the earliest opportunity.

However, data obtained from routinely collected data systems often require careful interpretation with respect to their quality, validity, timelines, bias, residual confounding and statistical stability [30]. For example, missing data are important sources of bias [31]. Routinely collected administrative data contain limited contextual information and represent an underestimate of total health outcomes for these individuals. Further, intervention and treatment for mental health conditions are not wholly captured across these data sources and may impact health outcomes.