Correlation between patient-determined inflammatory bowel disease severity and the inflammatory bowel disease severity index

Inflammatory bowel diseases (IBD) are relapsing inflammatory diseases of the intestine, which lead to debilitating symptoms and progressive intestinal damage over time.¹ Effectively assessing the impact of IBD on patients over time is important to determine the disease burden for an individual and across an IBD population. There are many domains by which to measure the impact of IBD on patients including disability (e.g., IBD Disability Index²), symptoms (e.g., Harvey-Bradshaw Index [HBI]³) and Simple Clinical Colitis Activity Index [SCCAI]⁴), quality of life (IBD questionnaire⁵), and composite scores of clinical symptoms as well as biomarkers (e.g., Crohn’s Disease Activity Index⁶). Although these and other measures provide useful cross-sectional data to describe how sick patients with IBD are at a given point in time, they do not reflect the cumulative burden and severity of disease.

The IBD Disease Severity Index (IBD-DSI) was developed in order to describe the severity of disease over the duration of time that a patient has been diagnosed with IBD.⁷ The methods for developing this score were meticulous and included a systematic review to find all possible components of disease activity and severity, panel discussion among IBD specialists and adaptive choice-based conjoint analysis. This type of conjoint analysis enables a list of variables to be presented within clinical scenarios to determine the importance of each for a clinician aiming to determine disease severity. Clinical scenarios are repeatedly presented until the variables can be consistently ordered by the group of clinicians undertaking the exercise. These variables were then weighted in the final index based on the level of significance determined through the process.

Since its publication, the IBD-DSI has been further studied to determine its validity by a number of groups. Swaminathan et al.⁸ prospectively followed 172 patients with IBD for 1 year and demonstrated that an elevated DSI was associated with an increased risk of depression, anxiety, perceived stress, reduced quality of life and a complicated IBD course. As part of a wider validation exercise, a multicentre study was undertaken to confirm inter-relator reliability and internal consistency.⁹ Construct validity was demonstrated with endoscopic indices, patient-reported outcomes measures and level of disability. The IBD-DSI accurately predicted a complicated disease course out to 24 months.⁹

Although the IBD-DSI was meticulously developed, it was done so by academic gastroenterologists with an interest in IBD and has not been evaluated or compared with patient perspectives of disease severity. Given the inherent subjectivity of the term “disease severity”, it is likely that there will be differences between patients and physicians regarding perspectives of IBD disease severity.

We aimed to correlate the physician-generated IBD-DSI with patient-perceived IBD severity. We also aimed to investigate whether individual factors modulate this correlation.

Methods

Setting

Three study centres (Christchurch Hospital, New Zealand, Dartmouth Hitchcock University Medical Center, United States of America [USA] and Northwestern University Medical Center, USA) took part in this cross-sectional study (ethics reference H19/027). Recruitment was undertaken between May 2019 and July 2021 and was delayed and curtailed by the COVID-19 pandemic.

Participants were eligible if they were aged 18 years and older, had IBD for 1 year or longer and were able to understand English fluently. Participants were recruited from outpatient clinics at the three centres.

Demographic and clinical data were extracted from the medical record including age, disease type, Montreal classification, gender, years since diagnosis, present medications and previous surgery.

The patient’s gastroenterologist completed the IBD-DSI score based on the medical record and patient interview. This index considers the following domains: a) the impact of the disease (i.e., symptoms and impact on daily activities), b) inflammatory burden (i.e., biomarkers and mucosal lesions) and c) disease course (i.e., how complicated the disease is, whether they respond to medication and what the disease extent is). Separate scoring systems for ulcerative colitis (UC; Table 1) and Crohn’s disease (CD; Table 2) are used. The IBD-DSI is scored 0–100, with 0 indicating the lowest severity possible and 100 the highest severity possible.

Variables

Patient-completed disease severity score (PCDSS)

For the PCDSS, participants were asked to rate the severity of their disease on a scale of 0–100 (Figure 1). Patients were asked to rate their disease severity based a composite measure of symptoms (e.g., abdominal pain, number of bowel motions per day), how IBD impacts their daily activities (e.g., impact on work and leisure), inflammatory burden (e.g., how bad results from blood tests and colonoscopies are) and their disease course (e.g., how many operations they have had and whether they have needed steroids before). These domains mirrored those in the IBD-DSI. Participants were given a linear scale from 0 to 100 and asked to mark on the scale how severe their disease was, with 0 reflecting no severity and 100 reflecting maximal severity. This scale was piloted on three patients with IBD for face validity before its use in this study.

View Figure 1–2, Table 1–6.

Other variables of interest

To measure gastrointestinal symptoms at recruitment, the HBI³ was completed by the treating physician for participants with CD and the SCCAI⁴ for participants with UC. Neuroticism was measured using the relevant questions from the self-reported Eysenck Personality Questionnaire Brief Version (EPQ-BV) using a five-point Likert scale.¹⁰ EPQ–BV has good internal consistency, test–retest reliability and concurrent validity.¹⁰

Coping strategies of participants were measured using the self-reported IBD-Cope, which is a concise IBD-specific coping strategy questionnaire that is reliable and valid.¹¹ It contains six questions, of which three interrogate “good” coping strategies and three interrogate “bad” coping strategies.

The four-item Pain Catastrophizing Scale (PCS) was used to measure pain catastrophising.¹² The PCS measures the negative emotional response to anticipated or actual pain¹³ and the questions are answered using a five-point Likert scale (ranging from not at all to all of the time). The four-item version correlates very highly with the original 13-item version (r=0.96).¹² Age, disease type, Montreal classification, gender, years since diagnosis, present medications and previous surgery were collected via the clinical records. The treating physician completed the IBD-DSI, HBI and SCCAI. The participant-completed PCDSS, neuroticism, coping and pain catastrophising were all self-reported by the participant via pen and paper.

Participants completed the PCDSS before the physician completed the IBD-DSI to ensure that participants came up with their score independently.

To help determine for which patients there was the greatest discrepancy between IBD-DSI and PCDSS, an “excess severity score” (ESS) was calculated by subtracting the physician-determined IBD-DSI from the PCDSS.

Study size

The initial power calculation determined that 300 participants would be needed across six sites to provide >80% power (two-tailed α=0.05) when correlation coefficients are >0.18. However, the COVID-19 pandemic halved the number of study centres. A recalculation determined a sample size of 150 participants across the three centres would enable correlations >0.23, to be detected as statistically significant (two-tailed α=0.05) with >80% power.

Ethical considerations

Ethical approval was obtained according to the institutional guidelines at each site.

Statistical methods

Statistical Packages for Social Sciences (SPSS) version 28 was used for analyses. For each discrete variable a frequency and percentage were calculated and for continuous variables a mean and standard deviation was calculated. An ESS was calculated by subtracting the physician-determined IBD-DSI from the PCDSS. For each variable, a general linear model was built with ESS as the dependent variable, IBD-DSI as a covariate and the variable of interest as a covariate for continuous or fixed for discrete. The file was split into UC and CD for analysing Montreal classifications and the clinical indices SCCAI and HBI. Interactions between predictors and diagnosis were also examined to see whether any predictors differed for UC vs CD. The alpha level was 0.05.

The histograms of the distributions of the IBD-DSI, PCDSS and ESS were examined and were found to be acceptably normal. Therefore, Pearson correlations were produced containing the IBD-DSI, PCDSS and ESS. A scatterplot was produced comparing IBD-DSI and PCDSS, with different coloured dots for centres, an x=y line and a regression line.

Finally, a stepwise regression was performed for predicting ESS based on IBD-DSI unconditionally and all other variables conditionally, with criteria being probability of F entry=0.05 and removal=0.10. The other variables included ones measured across CD and UC and not variables exclusive to CD or UC (i.e., HBI and Montreal classification for CD and SCCAI and disease extent for UC).

Data availability statement

The data underlying this article will be shared on reasonable request to the corresponding author.

Results

Participants

There were 170 participants approached from the three centres, with three having incomplete IBD-DSI data, leaving 167. Of these 167, three declined participation (98.2% response rate). There were 71, 46 and 47 participants recruited from Christchurch, Dartmouth Hitchcock and Northwestern, respectively. The average age was 41.6 years, 103 (62.8%) were female, and there were 98 participants (59.8%) with CD (Table 3).

Comparison of patient- and physician-estimated IBD severity

The ESS was determined by subtracting the IBD-DSI from the PCDSS. There were 164 complete IBD-DSI and therefore 164 calculatable ESSs. The mean ESS was 29.0 (Table 4), with a moderate positive correlation between the PCDSS and IBD-DSI (r=0.44, r²=0.19, p<0.001). Furthermore, there was a strong association between ESS and PCDSS (r=0.71, p<0.001) and a significant negative association between IBD-DSI and ESS (r=−0.32, p<0.001). The association between IBD-DSI and PCDSS is shown in Figure 2. As physician-determined severity gets higher, the ESS reduces. The regression equation reflecting this relationship is PCDSS=39.53+0.59*IBD-DSI.

Excess severity correlates

Variables that were positively associated with ESS at an alpha level of 0.10 or lower after IBD-DSI was accounted for were previous colectomy (p=0.03), any IBD- or bowel-related surgery (p=0.01) and UC extent of disease (p=0.049). Variables that trended insignificantly towards a positive association with ESS were resection (p=0.08), fistula/seton/abscess (p=0.07) and CD (p=0.06). 5-aminosalicylic acid (5-ASA) (p=0.01) was significantly negatively associated with ESS after controlling for IBD-DSI (Table 5). There was no difference in ESS between study centres (p=0.99).

Excess severity in UC vs CD

Significant interactions, signifying differing relationships between predictors and the ESS for UC and CD after controlling for IBD-DSI, were demonstrated for oral steroids (3.75 CD vs −28.52 UC, p=0.03) and bad coping (0.16 CD vs −0.21 UC, p=0.007). Colectomy (4.13 CD vs 32.38 UC, p=0.12), any IBD- or bowel-related surgery (3.89 CD vs 28.63 UC, p=0.099) and pain catastrophising (0.21 CD vs 0.004 UC, p=0.10) trended towards significance (Table 6).

5-ASA

Participants who took 5-ASA scored higher on the IBD-DSI (31.1 vs 24.7, p=0.17) but lower on the PCDSS (49.4 vs 56.5, p=0.16), meaning that the ESS was lower in those who were on a 5-ASA (p=0.003).

Multivariable analysis

A stepwise regression with the IBD-DSI being entered unconditionally revealed the significant variables were IBD-DSI (beta=−0.39, 95% confidence interval [CI]=−0.58–−0.21, p<0.001), any IBD- or bowel-related surgery (beta=8.03, 95% CI=0.80–15.27, p=0.03), and 5-ASA (beta=−9.67, 95% CI=−18.69–−0.65, p=0.04).

Discussion

Understanding the disease severity over a lifespan is essential to determine the impact on patients and communities. The IBD-DSI has been designed to achieve this through assessing a wide range of variables that may contribute to the IBD severity. A number of studies have shown that the IBD-DSI is a valid instrument that can also predict adverse disease outcomes out to 2 years.⁹ However, the IBD-DSI was not co-designed with patients and it is important to understand how well the IBD-DSI correlates with patients’ opinions of their own IBD severity. Furthermore, if the IBD-DSI does not correlate well, it is useful to understand the main drivers of such discrepancies.

In this study we have further extended the validation of the IBD-DSI by demonstrating a significant positive correlation between it and the PCDSS. Discrepancies regarding this correlation were greatest at lower disease severity with a mean 29.4-point ESS (patient-determined severity minus physician-defined severity). The equation: PCDSS=39.53+0.59*IBD-DSI accurately defines this relationship. This positive association evidences a level of concurrent validity for the PCDSS and IBD-DSI and is worthy of future study. While correlation does not imply causation, it is reasonable to infer that how severe a patient perceives their illness to be has some relationship to how severe an objective measure finds it to be, but future longitudinal studies with repeated measures would strengthen the argument that they are causally related.

In addition, variables that were positively associated with ESS included previous colectomy, any IBD- or bowel-related surgery and UC extent of disease. The only variable associated negatively with ESS was 5-ASA use. However, the present study has shown that the IBD-DSI and patients with IBD measure disease severity in a similar way. The stepwise multivariable model determined IBD-DSI, any IBD- or bowel-related surgery and 5-ASA use to be most strongly associated with ESS.

Overall, no psychological variables were associated with ESS. There were non-significant trends for good coping (p=0.15) and pain catastrophising (p=0.13) to be positively associated with ESS while neuroticism (p=0.37) and bad coping (p=0.94) had no significant association.

The reduction in ESS with increasing disease severity may be due to floor effects of determining disease severity for patients with less disease severity. These patients may concentrate on the few symptoms they have, driving up the severity score. Furthermore, regression to the mean may occur when participants in a study tend to score closer to the middle if they are not sure of their answer. Given that the mean IBD-DSI score for the study cohort was 25.8, regression to the mean could explain the larger ESS in those with less disease severity. It is important for clinicians to recognise that patients’ subjective experiences of severity may still be high even if they seem to have less severe IBD on objective measures; perhaps patients experience more severity than their treating team appreciates.

Another factor that could contribute to a wider ESS, which is specific to these questionnaires, may relate to patients scoring greater or lesser severity for a specific variable. For example, patients with a stoma may have different experiences leading them to over- or under-estimate the contribution of such a variable to the overall score. Furthermore, the IBD-DSI was not specifically designed for patients with UC who have had a colectomy. For example, resection or colectomy are not offered as variables in the IBD-DSI but would clearly reflect increased disease severity.

The only significant medication that was negatively correlated with ESS was 5-ASA. 5-ASA use was positively associated with an IBD-DSI and negatively with PCDSS. On the PCDSS side, it is possible that people on 5-ASA medications are aware that there are more potent medications for IBD and factor this into their score. On the IBD-DSI side it is possible some aspect of 5-ASA use is built into the scores. In addition, there was a non-significant trend towards steroids being negatively associated with ESS and it is likely that those on steroids are in steroid-induced remission and so are actually feeling well. There was also an interesting interaction effect in that oral steroids were a negative predictor of ESS in UC, more so than CD, although there were small numbers in each diagnosis group.

It is possible participants were subject to the Hawthorne effect, wherein they answered differently because they knew their answers would be observed.¹⁴ Participants were recruited from specialist gastroenterology outpatient clinics and it is possible this affects their beliefs about their disease severity, especially when one considers that many patients with IBD are not under the care of a specialist.^15–17

This study under-recruited due to the COVID-19 pandemic and so there were limitations in the power of the study, especially with regards to the psychological variables. Nevertheless, some significant associations were found. Moreover, some potentially important variables were not collected, such as dose and duration of medications as well as comorbidities. Future studies should ideally collect these datapoints.

The PCDSS was used for the first time in this study but was shown to correlate with the IBD-DSI and so seems to have measured what it purports to measure. It is intended as a subjective scale, not unlike the 0–100 visual analogue scale used in the EQ-5D-5L questionnaire.^18–22 Because it is intended as a patient-defined and subjective measure of disease, it should not be expected to correlate perfectly with the IBD-DSI, which was indeed a finding of this study. It does mention symptoms (e.g., abdominal pain, number of daily bowel motions), impact on daily activities (e.g., work and leisure), inflammatory burden (blood test and colonoscopy results), and disease course (e.g., previous operations or steroids) and so is comprehensive while being partly subjective.

The calculation of ESS may seem unusual given the apparent heterogeneity of IBD-DSI and PCDSS in terms of what they purport to measure. However, this was necessary to identify variables with the greatest discrepancy between IBD-DSI and PCDSS, and where patients assigned different weights than in the gastroenterologist-developed IBD-DSI.

This study recruited from New Zealand and the USA, and excess severity was the same across the centres despite the different subcultures, economies and health systems. This suggests both instruments are generalisable across continents.

A similar study from Germany²³ that compared 89 pairs of the physician-completed IBD-DSI to a patient global self-assessment and visual analogue scale also reported moderate agreement between physicians and patients. Furthermore, patients tended to rank their disease severity higher than physicians in the German study as well. Given that there are now two studies in different languages and using different scales reporting similar associations, it is likely that the IBD-DSI is valid. The moderate correlations reported in both studies reflect the comparisons between different constructs, namely objective severity and subjective patient experience.

Conclusions

There is a moderate positive association between PCDSS and IBD-DSI but with greater discrepancies at the lower levels of IBD-DSI. The CD and UC IBD-DSI are differently calculated and so there were some differences between CD and UC in terms of ESS as well as what predicted ESS. Surgical procedures especially predicted ESS, at least in part driven by surgical procedures not being included in the IBD-DSI for UC, demonstrating the IBD-DSI is not valid for UC patients who have had a colectomy.