The Use of, and Outcomes for, Inflammatory Bowel Disease Services During the Covid-19 Pandemic

A Nationwide Observational Study

Mohammed Deputy; Kapil Sahnan; Guy Worley; Komal Patel; Violeta Balinskaite; Alex Bottle; Paul Aylin; Elaine M Burns; Ailsa Hart; Omar Faiz


Aliment Pharmacol Ther. 2022;55(7):836-846. 

In This Article


Study Design and Data Source

A nationwide observational study of IBD services was performed using Hospital Episode Statistics (HES) Admitted Patient Care database. The use of HES in research has been described previously.[4] The HES data are the routinely collected administrative inpatient healthcare data for all National Health Service (NHS) patients in England, treated in both NHS and private hospitals. The data include patient demographic and socioeconomic data, coded diagnostic and procedural data, and outcome data such as mortality and length of stay.[4]

Study Population and Time-span

The study population was patients admitted in England between 1 January 2015, and 31 January 2021, aged 18 years and older, and given a coded diagnosis of IBD in keeping with a previous or new diagnosis of IBD. The epidemic was defined as beginning 1 February 2020 to ensure we captured any difference in presentations or outcomes in the period before the national lockdown was announced by the UK government on 23 March 2020.

Identification of IBD Services

We investigated two types of emergency IBD admission and eight emergency and elective procedures for IBD. These are outlined in Table 1.

Emergency IBD admissions were identified by World Health Organisation International Classification of Disease 10th Revision (ICD-10) codes in the primary diagnosis field admitted as an emergency. IBD procedures were identified by Office of Population Censuses and Surveys Classification of Interventions and Procedures Version 4 (OPCS-4) code in any procedure field with a diagnosis of IBD in any diagnosis field. The specific codes used in each analysis can be found in Supplementary Table S1.


Mortality was defined as 30-day all-cause in-hospital mortality. 30-day mortality was evaluated up until February 2021. Readmission was defined as readmission within 28 days of discharge to any NHS hospital for any reason. Length of stay was calculated as the difference between the admission date and discharge date in days.

Statistical Analysis

Patient characteristics included for analysis were age, sex, ethnicity, Charlson comorbidity score with weights tailored to HES, and social deprivation quintile. Patients with invalid data recorded for age or sex were excluded. Ethnicity was considered in six major groups: White; Mixed; Asian; Black; Chinese or Other; and Not Known/Not Stated. Comorbidity was categorised as a Charlson score of 0–2 and those with a score of 3 or more. The Charlson score was calculated from secondary diagnosis codes. Social deprivation was categorised into population-weighted quintiles using the Carstairs index, with 1 being the least deprived and 5 being the most deprived (6 represents not assigned).

A concomitant SARS-CoV-2 diagnosis was identified, in any secondary diagnostic field, within the same episode, by the emergency ICD-10 codes: U071 (SARS-CoV-2 infection confirmed by laboratory testing) or U072 (clinical or epidemiological SARS-CoV-2 infection where laboratory confirmation is inconclusive or not available).

Admissions and procedures performed each month were counted. To forecast the counterfactual number of cases that would have happened without the covid-19 pandemic, historical trends for the previous 5 years were plotted with a local smoother to assess for overall trend and seasonality.

Autoregressive integrated moving average (ARIMA) models were run to forecast the counterfactual with the proc arima statement in SAS.[5] ARIMA models can be used to predict population-level changes and are more suitable than standard regression analysis, which assumes a time series is not autocorrelated.[6] ARIMA models have been used previously to forecast demand for hospital services.[7]

An ARIMA model consists of three parameters: p, d and q. p refers to the autoregressive (AR) part of the model, d refers to the degree of differencing (I) and q refers to the order of moving average (MA) part of the model.

An autoregressive (AR) model is where the forecasted variable is predicted by one or more observed lagged values. A moving average model is where the predicted variable is predicted by one or more observed lagged values of the error. Differencing is performed to make a non-stationary data series stationary (ie, it removes the pre-existing trend), which is a requirement for an ARIMA model. The method used to build the ARIMA models is detailed in Appendix 1.

The models were used to forecast the counterfactual (the expected number based on pre-pandemic levels) for 12 months and then compared with the observed counts to calculate deficits. Relative deficits were calculated by dividing absolute deficits by the forecast volume and multiplying by 100.

Proportions were compared with the chi-squared test and medians were compared with the Mann-Whitney test.

All statistical analyses were performed in SAS version 9.4.

Sensitivity Analysis

A sensitivity analysis was conducted where emergency admissions with a code for ulcerative colitis or Crohn's disease in any diagnosis position were included. ARIMA models were constructed for these analyses, and the deficit based on pre-pandemic trends in admissions was calculated.

Ethics Statement

We have approval from the Secretary of State and the Health Research Authority under Regulation 5 of the Health Service (Control of Patient Information) Regulations 2002 to hold confidential data and analyse them for research purposes (CAG ref 15/CAG/0005). We have the approval to use them for research and measuring the quality of delivery of healthcare, from the London—South East Ethics Committee (REC ref 20/LO/0611).