key: cord-0811047-sdqcgl7o
authors: Carr, Matthew J; Wright, Alison K; Leelarathna, Lalantha; Thabit, Hood; Milne, Nicola; Kanumilli, Naresh; Ashcroft, Darren M; Rutter, Martin K
title: Impact of COVID-19 on diagnoses, monitoring, and mortality in people with type 2 diabetes in the UK
date: 2021-05-11
journal: Lancet Diabetes Endocrinol
DOI: 10.1016/s2213-8587(21)00116-9
sha: 97c371bdc67f871a1a6cf6d9b039fb5434dd2c15
doc_id: 811047
cord_uid: sdqcgl7o

nan

We conducted a retrospective cohort study using primary care electronic health A total of 24,440,354 patients were included for estimating the expected rates in the pre-COVID-19 period (January 2010 to February 2020). The CPRD contains anonymised consultation records and includes patient demographic information, symptoms, diagnoses, medication prescriptions, and date of death. We also examined practice-level Index of Multiple Deprivation (IMD) quintiles [3] , a measure representing an area's relative level of deprivation, ranked within each UK nation.

To enable comparisons of rates before and after the start of the COVID-19 outbreak, we included patient records from January 2010 that established long-term trends and patterns of seasonality. We focussed primarily on reporting observed versus expected rates from 1 st March 2020 to 10 th December 2020. First, we estimated incidence rates of T2D diagnoses, new prescriptions for metformin (the most commonly prescribed medication in new-onset T2D) and insulin, and rates of HbA1c testing and mortality in people with T2D.

and Aurum (see https://clinicalcodes.rss.mhs.man.ac.uk). In line with guidance from the CPRD's central administration, data from the Aurum and GOLD databases were analysed separately, with data from Aurum restricted to English practices and GOLD providing information on practices in Northern Ireland, Scotland and Wales. The use of two discrete data sources also enabled independent replication of within-study prospective changes in study outcomes. All code lists and medication lists were verified by two senior clinical academics (a diabetologist: MKR, and a senior academic pharmacist: DMA).

For each patient, we defined a 'period of eligibility' for study inclusion which commenced on the latest of: the study start date (1st January 2010); the patient's most recent registration with their practice; the date on which data from the practice was deemed to be 'up-to-standard' by the CPRD. A patient's period of eligibility ended on the earliest of: registration termination; the end of data collection from their practice; death. For incident diagnoses and prescriptions, we also applied a 'lookback' period during which a patient was required to have been registered for at least a year prior to the event. Flow diagrams illustrating the delineation of the study cohorts using CPRD Aurum and GOLD are presented in Supplemental Figures in https://www.medrxiv.org/content/10.1101/2020.10.25.20200675v2. The denominator for the incidence rates was the aggregate person-months at risk for the whole eligible study population. Mortality and testing rates in people with T2D were calculated using the person-months at risk from all those with T2D as the denominator. Incidence, mortality and testing rates were stratified by sex, age group (<18, 18-29, 30-44, 45-64, 65-79 and ≥80 years), practice-level deprivation (IMD quintiles) and region (in England) or nation (in the rest of the UK).

The data were structured in a time-series format with event counts and 'personmonths at risk' aggregated (by year and month) with stratification by sex, age group, deprivation quintile and region (or nation in GOLD). Mean-dispersion negative binomial regression models were used to establish long-term trends and patterns of seasonality using monthly data covering January 2010 to February 2020. We selected negative binomial regression models over Poisson models because of the high variation in the outcomes studied. The denominator for the incidence rates was the aggregate person-months at risk for the whole eligible study population. Mortality and testing rates were calculated using the person-months at risk from all patients with T2D as the denominator. We then used the regression models to forecast expected monthly event counts from March 2020 onward. The natural logarithm of the denominator (person-months at risk) was used as an offset in each regression model. To account for possible seasonality and long-term linear trends, calendar month was fitted as a categorical variable and time as a continuous variable with the number of months since the start of the study serving as the unit of measurement.

The monthly expected rates, and their 95% confidence intervals, were plotted against the observed rates. As they share a common denominator, differences between expected and observed monthly rates are expressed as a percentage 'rate reduction (or increase)'.

Extrapolated estimates of the number of missed (or delayed) diagnoses of T2D were derived using the discrepancy between observed and expected frequencies from March 2020 onward, and approximations of the proportional representation of the populations of England and the rest of the UK (in CPRD Aurum and GOLD respectively) using data from the Office for National Statistics [4] .

All data processing and statistical analyses were conducted using Stata version 16 (StataCorp LP, College Station, TX, USA). We followed RECORD (REporting of studies Conducted using Observational Routinely-collected health Data) guidance [5] .

Comparison of observed and expected monthly incidence rates for type 2 diabetes in primary care, HbA 1c monitoring in type 2 diabetes, new prescriptions for metformin and insulin, and deaths in people with type 2 diabetes in England (CPRD Aurum) during the first COVID-19 peak in April 2020 and overall between March 1 and December 10, 2020. 

Data Resource Profile: Clinical Practice Research Datalink (CPRD)

Data resource profile: Clinical Practice Research Datalink (CPRD) Aurum

Approach to record linkage of primary care data from Clinical Practice Research Datalink to other health-related patient data: overview and implications

Population estimates

The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement

Translational Research Centre. The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care. The funding source had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all of the data and the final responsibility to submit for publication.

A preliminary version of this article was deposited (prior to peer review) in the medRxiv preprint repository: https://www.medrxiv.org/content/10.1101/2020.10.25.20200675v2