Patient outcomes and healthcare utilisation following implementation of an advanced therapy care pathway for inflammatory bowel disease: a multicentre, prospective quasi-experimental preimplementation/postimplementation study in the Netherlands

Abstract

Objective

Care pathways have the potential to decrease treatment variation and improve patient outcomes. This study aims to evaluate the effect of an advanced therapy care pathway (ACP) on patient outcomes, healthcare utilisation and treatment variation among patients with inflammatory bowel disease (IBD) in the Netherlands.

Methods

A multicentre, prospective quasi-experimental preimplementation/postimplementation study was conducted, with a preimplementation (December 2020–December 2021) and postimplementation period (March 2022–March 2023). Outcomes were collected from electronic medical records and validated questionnaires according to the International Consortium for Health Outcomes Measurement standard set. The effect of the ACP was analysed using difference-in-differences (DiD) analysis.

Results

1173 patients were included; the majority was female (55%) with a median age of 45 years (IQR 33–58 years). Most patients had Crohn’s disease (64%). Patients reported a high level of disease control and this did not change postimplementation (DiD-estimate: 0.12, 95% CI −0.74 to 1.01). Number of emergency room visits and hospital admissions decreased significantly 6 months postimplementation; however, the effect attributable to the ACP was only maintained for admissions (DiD-estimate visits: −0.05, 95% CI −0.10 to 0.01; DiD-estimate admissions: −0.06, 95% CI −0.11 to −0.03).

Conclusion

Implementation of an ACP, in which a minimum of care is defined, can achieve outcomes comparable to standard practice, highlighting the value of prioritising structured care pathways, especially given the current burden of IBD on healthcare systems. An ACP can guide healthcare providers and policymakers in standardising treatment and improving quality of care in chronic disease management.

Study registration

NL8276/NL-OMON21751 (https://onderzoekmetmensen.nl/en/trial/21751).

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Inflammatory bowel disease (IBD) imposes a substantial economic burden, driven by high healthcare utilisation and rising costs of advanced therapies. Prior care pathway studies in IBD showed promising reductions in healthcare utilisation but were limited by small, retrospective, single-centre designs without causal inference.

WHAT THIS STUDY ADDS

• This is the first multicentre prospective study to causally evaluate an advanced therapy care pathway (ACP) for IBD, finding significant reductions in emergency room visits and hospital admissions 6 months postimplementation. The ACP, defined as a minimum standard of care, achieved these reductions without compromising disease control, remission or quality of life.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• A structured minimum standard of care for patients with IBD on advanced therapies can reduce unplanned healthcare utilisation without compromising outcomes, supporting wider pathway adoption in clinical practice. For policymakers, care pathways offer a practical, scalable tool for standardising IBD treatment and managing its growing economic burden.

Introduction

Inflammatory bowel diseases (IBDs), including Crohn’s disease (CD) and ulcerative colitis (UC) are chronic, relapsing–remitting diseases that impose a substantial and growing economic burden on healthcare systems worldwide. In Europe, annual direct healthcare costs of IBD are estimated to exceed €5.6 billion, with an additional €4.6 billion in indirect costs such as productivity loss and long-term work disability.¹ In the USA, the estimated annual cost of IBD approximates US$50 billion, including US$23.8 billion in direct medical expenses.^2–4 This financial burden is expected to rise further due to increasing incidence rates, increasing costs for treatment and persistent treatment variation in clinical practice.^{1 5}

Key drivers of this economic burden are high healthcare utilisation among patients with IBD and increasing medication expenditure.⁶ Patients require frequent diagnostic imaging, laboratory monitoring, endoscopic evaluations and long-term treatment with biologics or small molecules.^7–10 Despite therapeutic advances, up to 50% of patients are hospitalised at least once within 10 years of diagnosis, and approximately 30% of patients with CD undergo surgical resection.¹¹ Compounding these challenges is the variation in how IBD is managed across healthcare providers. Treatment variation, defined as inconsistent approaches to diagnosis, monitoring and therapy, can result in suboptimal outcomes, such as increased emergency department visits, prolonged hospitalisations and overuse of diagnostic tests.^12–14 These inconsistencies underscore the importance of developing effective strategies to optimise patient care while efficiently managing healthcare resources.

One potential strategy is the implementation of care pathways. Care pathways are practical tools for aligning clinical practice with evidence-based standards and reducing unnecessary variation.^{12 13 15 16} By ensuring uniform practices, care pathways help to allocate resources where they are most needed, thereby reducing the overuse of hospital care and potentially reducing healthcare costs.¹⁵ For example, protocolised care pathways in emergency abdominal surgery are associated with a shorter hospital stay and a reduction in postoperative complications.¹⁷ In IBD care, care pathways have been developed and showed promising effects on anaemia prevalence, emergency department utilisation and number of performed upper endoscopies and imaging studies.^18–20 However, these studies were limited by small, retrospective designs and single-centre implementation. They focused solely on specific complications of IBD or only reported outcomes on healthcare utilisation.

We conducted the multicentre prospective IBD Value Study to improve IBD care in the Netherlands by implementing an advanced therapy care pathway (ACP) for the treatment of patients with IBD using biologics and small molecules.²¹ The ACP standardises prescribing, switching and monitoring advanced therapies. It was hypothesised that implementing an ACP would improve patient outcomes (eg, remission, quality of life) and reduce healthcare utilisation, while decreasing treatment variation across hospitals. This study aimed to evaluate the causal effect of an ACP implemented in the Netherlands on patient outcomes and healthcare utilisation, as well as to assess its effect on treatment variation across hospitals.

Methods

Design and study timeline

We conducted a multicentre, quasi-experimental preimplementation/postimplementation study with concurrent controls across eight hospitals in Southwest Netherlands (December 2020–March 2023).²¹ All hospitals are members of IBD BeterKeten, an ongoing collaboration to improve quality of care for patients with IBD. One of the hospitals is an academic centre, five are large teaching centres and two are community hospitals. The patient population of the hospitals is diverse, with a range of ethnic, socioeconomic and educational backgrounds.²²

The study had a total observation period of 27 months. The first 12 months were the preimplementation period. During this period, healthcare providers delivered care according to current guidelines and local protocols, determining monitoring frequency, procedures and timing for therapy adjustments. After the preimplementation period, the ACP was implemented in six of the eight participating hospitals over a 3-month period. The other two hospitals were used as natural controls, the selection of control hospitals was based on pragmatic considerations, and they were not involved in the design process of the ACP. Following implementation, patient outcomes and healthcare utilisation were assessed over a subsequent 12-month post-implementation period.

Advanced therapy care pathway

The ACP provided standardised guidelines for managing patients with IBD using advanced therapies, based on national and international guidelines.^23–28 The ACP was developed by a multidisciplinary and multicentre working group, guided by international and national guidelines, available scientific evidence, current best practices, and expert opinions in the absence of scientific evidence.

The ACP was designed to provide a minimum of care that should be given to patients with IBD and can be adapted by healthcare providers based on individual patient needs. It includes systematic flare assessment, structured treatment algorithms, annual review of disease activity and colorectal cancer surveillance. For each approved medical treatment (eg, infliximab, adalimumab, golimumab, vedolizumab, ustekinumab, tofacitinib, filgotinib and ozanimod), the ACP outlines standardised induction and maintenance protocols with predefined clinical and laboratory evaluations (eg, blood samples at week 0, 6 and 14 during the start of medication), safety monitoring and timing of follow-up visits (two per year after more than 1 year on therapy). While the exact format of consultations (in-person, telephone or digital) may vary between centres, the sequence and content of care were uniform across all intervention hospitals. An overview of the differences and similarities between the European Crohn’s and Colitis Organisation (ECCO) guidelines and the ACP can be found in online supplemental table 1.

Population

The eligible population consisted of adult patients (≥18 years) diagnosed with IBD for at least 3 months and receiving treatment at one of the eight participating hospitals. Patients had to be treated or starting treatment with advanced therapies, including, but not limited to, infliximab, adalimumab, golimumab, vedolizumab, ustekinumab and tofacitinib, as well as any new treatments registered during the study period. Exclusion criteria were insufficient knowledge of the Dutch language and lack of internet access to complete questionnaires. Eligible patients were identified with help of the business intelligence department at the start of the study. These patients received an information letter or email from their hospital, asking for participation in the study. Patients could also provide informed consent to their treating IBD specialist or nurse during outpatient visits or hospital admissions. Patients were considered non-participants if they actively declined participation or did not respond to the invitation. Independent of enrolment date, outcomes were collected over the same period (December 2020 until March 2023).

Outcomes

Patient outcomes and healthcare utilisation were measured using the standard set defined by the International Consortium of Health Outcomes Measurement (ICHOM), focusing on outcomes most meaningful to patients.²⁹ The primary outcome was patient-reported disease control, measured using the IBD-Control-8 score, a validated questionnaire to measure disease control from a patient’s perspective.³⁰ Secondary outcomes from the ICHOM standard set included presence of anaemia or active fistulas, frequency of emergency room (ER) and outpatient visits, number and length of hospital admissions, IBD-intervention-related complications, long-term steroid use (defined as cumulative use exceeding 3 months a year) and body mass index (BMI).²⁹ Remission was defined as biochemical, endoscopic/radiological, clinician-reported and patient-reported remission. Biochemical remission was defined as a faecal calprotectin below 100 µg/g for CD and below 250 µg/g for UC.^{31 32} Endoscopic/radiological remission required no visible inflammation on endoscopy or imaging. Clinician-reported remission was recorded when the healthcare provider documented on remission in the electronic health record. Patient-reported remission was assessed using the Manitoba IBD Index (MIBDI).³³ Generic health-related quality of life, measured with the 10-item Patient-Reported Outcomes Measurement Information System Global Health (PROMIS-10) questionnaire and health-related quality of life measured with the EuroQol Five Dimensions Five Levels (EQ-5D-5L) were added to the ICHOM outcomes.^{34 35} Questionnaires to assess patient-reported outcomes were sent digitally every 6 months; a further detailed overview of the timing can be found in our protocol.²¹ Administrative data were automatically collected from the electronic medical records and linked to the survey results using specifically developed software.³⁶ Outcome definitions can be found in online supplemental table 2.

To account for case-mix differences among patients from the eight different hospitals, the ICHOM set of case-mix variables was collected. These variables included age, sex, education level (as defined by UNESCO³⁷), smoking status, comorbidities, diagnosis, Montreal classification, extraintestinal manifestations and specific conditions such as primary sclerosing cholangitis, hepatitis B, tuberculosis and HIV infection. Most variables were extracted from the electronic medical records, with comorbidities assessed using the Self-Administered Comorbidity Questionnaires (SCQ).³⁸ Surveys to collect case-mix variables were sent out at inclusion, after the implementation period and at the end of the study.

Statistical analysis

An analytic sample size calculation was not conducted due to the longitudinal clustering of data. Instead, simulations were performed and indicated that 25 patients per hospital had to be included to have sufficient power to identify a clinically meaningful change in the IBD-Control score of 1.²¹ The eligible population consisted of approximately 4000 patients with IBD who were treated with advanced therapies in one of the participating hospitals.

Missing data were handled using single imputation nested within a bootstrap procedure.³⁹ The analysis used a difference-in-differences (DiD) approach to estimate the causal effect of the use of the ACP on patient outcomes and healthcare utilisation, while rigorously correcting for observed time trends.²¹ The DiD-approach is commonly used to analyse the effect of an intervention using longitudinal data, assuming that differences between control and intervention hospitals do not change over time and that factors that change over time affect both groups equally.⁴⁰ It compares how much the outcomes changed over time in the intervention hospitals, relative to the change in control hospitals, without adding covariates to adjust for differences. For this comparison to be valid, the parallel trends assumption needs to be satisfied.^40–42

Pretrend assessment

To assess the parallel trend assumption, average outcomes at 6-months and 12-months preimplementation were calculated and visualised. The parallel trend assumption was satisfied when a comparable trend was visible between control and intervention hospitals. This was formally tested using the DiD package in R, by assessing if the 95% CIs for the preintervention estimate included 0.⁴³ For the outcomes length of stay and ER visits, the parallel trend assumption was not satisfied. For these outcomes, we estimated inverse probability of treatment weights using propensity scores. First, we identified per variable which baseline variables were associated with preintervention changes in the outcomes using a linear regression model. Variables with a pooled p value below 0.05 were identified as significant predictors of preintervention change and subsequently used as covariates to estimate the propensity scores. Secondary outcome p values were adjusted using the Benjamini-Hochberg procedure to control for the false discovery rate. For both outcomes, no significant predictors were found and propensity score weighting was not applied. Consequently, the DiD-analysis was conducted with an adjustment for the preimplementation change scores. The preimplementation change scores were defined as the difference between the 12-month and 6-month measurement of the outcome and calculated separately for each individual patient. This pretreatment change score was then included as a covariate in the DiD-analysis.

DiD-analysis

The DiD analysis was implemented using the group-time average treatment effect estimator, as follows:

$${\displaystyle {\displaystyle \begin{array}{ll}ATT\left(g,t\right)& =E\left[Yt\left(g\right)-Y_{\left\{g-1\right\}}\left(g\right)|Gg=1\right]\\ -E\left[YT\left(0\right)-Y_{\left\{g-1\right\}}\left(O\right)|C=1\right],\end{array}}}$$

where g denotes the period in which a hospital first implemented the ACP, t denotes the outcome measurement period, Gg=1 indicates hospitals first treated in period g and C=1 indicates hospitals in the comparison group. Hospitals that never implemented the ACP during the study period were coded as ‘never treated’ and served as the comparison group. Intervention hospitals were coded as first treated at 21 months, corresponding to ACP implementation. ATT(g,t) thus represents the average effect of the ACP on the outcome at time t, among hospitals that adopted the ACP at time g, relative to the counterfactual change observed in never-treated hospitals over the same interval.

Variation

To evaluate the effect of the ACP on variation between hospitals, data from the six intervention hospitals were analysed for the preimplementation and postimplementation periods using generalised mixed models and intraclass correlation coefficients (ICCs) were calculated.

For each outcome, two models were fitted across the imputed datasets. The baseline model included a fixed effect for period and case-mix variables. Random effects were a patient-specific random intercept to account for repeated measurements and patient-level variability, and a hospital-specific random intercept to capture heterogeneity between hospitals over the study period. The interaction model extended the baseline model by replacing the hospital-specific random intercept with a hospital-by-treatment interaction random effect. This term allowed for the effect of the implementation of the ACP on the outcomes to vary across different hospitals, thereby assessing changes in the pattern of between-hospital variation post-implementation.

ICCs were calculated for each random effect to quantify the proportion of total variance explained by that level. ICCs were pooled across the imputed datasets. A high ICC indicates that a larger proportion of the observed variation in outcomes is associated with the treating hospital.⁴⁴ Model comparison was performed using a Likelihood Ratio Test to assess whether the more complex interaction model provided a significantly better fit than the baseline model. P-values below 0.05 were considered statistically significant.

All statistical analyses were performed using R V.2024.04.2+764. Imputed datasets were generated with mice and custom scripts to summarise and integrate results from each imputed dataset into the final analysis.⁴⁵ The DiD package was used to calculate the average treatment effect.⁴³ For model fitting the glmmTMB package was used.⁴⁶

All authors had access to the study data and reviewed and approved the final manuscript. The SQUIRE 2.0 checklist was used as a framework for reporting the outcomes of this study (online supplemental file 2).⁴⁷

Results

Of the 4692 eligible patients, 1173 (25%) were included. Of the remaining 3159 patients, 333 (7%) actively declined participation and 3186 (68%) did not respond to the invitation. Of the 1173 included patients, 985 (84%) were enrolled during the preimplementation period and 188 (16%) during the subsequent implementation and postimplementation period. Nine patients were lost to follow-up (death, n=9). Patients in intervention and control hospitals had comparable baseline characteristics in most aspects (table 1). Patients in intervention hospitals, however, reported higher self-reported comorbidities scores and a higher proportion had rheumatological extraintestinal manifestations. Furthermore, a larger number of patients in the intervention hospitals had previously received treatment with infliximab or adalimumab. Baseline characteristics per hospital can be found in online supplemental table 3.

Table 1. Baseline characteristics.

Characteristic	Intervention hospitals (n=841)	Control hospitals (n=332)	P value
Demographics*
Age	45 (33, 58)	47 (33, 60)	0.26
Female	473 (56%)	172 (52%)	0.18
Education level			0.24
Lower	200 (27%)	75 (25%)
Middle	277 (38%)	103 (34%)
Higher	260 (35%)	123 (41%)
Smoking status			0.16
Never	315 (46%)	111 (40%)
Ex	251 (37%)	120 (43%)
Current	118 (17%)	48 (17%)
Comorbidities*
SCQ	1.00 (0.00, 3.00)	1.00 (0.00, 2.50)	0.21
History of TBC	9 (1.3%)	3 (1.1%)	1.00
History of Hepatitis B	12 (1.8%)	2 (0.7%)	0.37
HIV/AIDS	2 (0.3%)	0 (0%)	1.00
PSC	14 (2.1%)	3 (1.1%)	0.42
Diagnosis*
Crohn’s disease	543 (65%)	209 (63%)	0.61
Ulcerative colitis	279 (33%)	117 (35%)	0.55
IBD-unknown	16 (1.9%)	6 (1.8%)	1.00
Montreal classification*
Disease extent			0.36
E1	14 (1.7%)	4 (1.2%)
E2	93 (11%)	30 (9.0%)
E3	188 (22%)	89 (27%)
Age at diagnosis			0.05
A1	51 (6.1%)	12 (3.6%)
A2	384 (46%)	138 (42%)
A3	108 (13%)	59 (18%)
Disease location			0.61
L1	131 (16%)	54 (16%)
L2	114 (14%)	51 (15%)
L3	296 (35%)	102 (31%)
L4 (isolated)	2 (0.2%)	1 (0.3%)
Concomitant proximal small bowel disease	71 (8.5%)	20 (6.0%)	0.35
Disease behaviour			0.08
B1	336 (40%)	147 (44%)
B2	119 (14%)	29 (8.7%)
B3	87 (10%)	33 (9.9%)
Perianal disease	148 (18%)	49 (15%)	0.50
Extraintestinal manifestations*
History of dermatological EIMs	60 (7.1%)	15 (4.5%)	0.12
History of rheumatological EIMs	133 (16%)	18 (5.4%)	<0.01
History of hepatobiliary EIMs	17 (2.0%)	2 (0.6%)	0.12
History of opthalmological EIMs	32 (3.8%)	8 (2.4%)	0.29
Treatment history*
Previous therapy with infliximab	294 (35%)	91 (27%)	0.01
Previous therapy with adalimumab	203 (24%)	58 (18%)	0.02
Previous therapy with golimumab	10 (1.2%)	7 (2.1%)	0.27
Previous therapy with vedolizumab	54 (6.4%)	24 (7.3%)	0.68
Previous therapy with ustekinumab	31 (3.7%)	9 (2.7%)	0.49
Previous therapy with tofacitinib	10 (1.2%)	9 (2.7%)	0.07
History of IBD-related surgery	279 (33%)	98 (30%)	0.24
Treatment at baseline*
Infliximab	272 (32%)	108 (33%)
Adalimumab	226 (27%)	77 (23%)
Golimumab	6 (0.7%)	1 (0.3%)
Vedolizumab	124 (15%)	49 (15%)
Ustekinumab	79 (9.4%)	45 (14%)
Tofacitinib	14 (1.7%)	5 (1.5%)
No biologics	120 (14%)	47 (14%)

^*Median (IQR) or n (%).

EIM, extraintestinal manifestations; IBD, inflammatory bowel disease; PSC, primary sclerosing cholangitis; SCQ, Self-administered Comorbidity Questionnaire; TBC, tuberculosis.

Outcomes are grouped by primary outcomes (IBD-control), remission status, healthcare utilisation, symptoms, function and quality of life. For each outcome, the mean values for the preimplementation and postimplementation periods are shown, followed by the DiD-estimate representing the causal effect of the ACP at the end of the study period.

IBD-control

Patients in the intervention and control hospitals reported a high level of disease control (figure 1A and table 2). The implementation of the ACP had no significant effect on patient-reported disease control after 1 year (0.12, 95% CI −0.74 to 1.01), figure 1B).

Figure 1. IBD-control scores per group over time (A) and DiD-analysis (B). ATT, average treatment effect on the treated; DiD, difference-in-differences; IBD, inflammatory bowel disease.

Table 2. Remission outcomes and ICC by hospital type preimplementation and postimplementation.

	Missing data, %	Hospital type	Period	Mean (95% CI)	ICC, % (p value)	DiD estimate (95% CI)
IBD-control
	40	Control	Preimplementation	12.1 (10.3 to 13.2)
			Postimplementation	12.3 (10.1 to 13.3)
		Intervention	Preimplementation	11.8 (9.76 to 12.9)	0.19
			Postimplementation	11.9 (9.57 to 13.0)	0.50 (1)	0.12 (−0.74 to 1.01)
Remission status
Biochemical remission	41	Control	Preimplementation	0.56 (0.51 to 0.62)
			Postimplementation	0.63 (0.52 to 0.68)
		Intervention	Preimplementation	0.54 (0.50 to 0.58)
			Postimplementation	0.64 (0.55 to 0.69)	0.42
				11.9 (9.57 to 13.0)	0.82 (1)	0.04 (−0.04 to 0.10)
Endoscopic/radiological remission	79	Control	Preimplementation	0.43 (0.21 to 0.65)
			Postimplementation	0.50 (0.06 to 0.91)
		Intervention	Preimplementation	0.57 (0.25 to 0.66)	0.34
			Postimplementation	0.49 (0.08 to 0.89)	0.61 (1)	−0.04 (−0.16 to 0.06)
Clinician-reported remission	1	Control	Preimplementation	0.81 (0.77 to 0.84)
			Postimplementation	0.86 (0.83 to 0.90)
		Intervention	Preimplementation	0.79 (0.77 to 0.82)	0.00
			Postimplementation	0.83 (0.81 to 0.86)	0.85 (1)	−0.02 (−0.08 to 0.03)
Patient-reported remission (MIBDI)	35	Control	Preimplementation	0.44 (0.35 to 0.55)
			Postimplementation	0.47 (0.39 to 0.55)
		Intervention	Preimplementation	0.38 (0.28 to 0.50)	0.33
			Postimplementation	0.45 (0.37 to 0.55)	0.17 (1)	0.05 (−0.04 to 0.12)

DiD, difference-in-differences; IBD, inflammatory bowel disease; ICC, intraclass correlation coefficient.

Remission status

Rates indicating the remission status of a patient with IBD were stable across all groups, with small, non-significant increases in biochemical and patient-reported remission postimplementation (table 2). Exploratory analyses indicated that the ACP was not significantly associated with changes in biochemical (0.04, 95% CI −0.04 to 0.10), endoscopic/radiological (−0.04, 95% CI −0.16 to 0.06), clinician-reported (−0.02, 95% CI −0.08 to 0.03) and patient-reported remission (0.05, 95% CI −0.04 to 0.12) (table 2, online supplemental figures 1–4).

Healthcare utilisation

Although the number of ER visits and hospital admissions per patient decreased numerically in intervention hospitals, the ACP’s significant effect was not sustained over time for ER visits (−0.05, 95% CI −0.10 to 0.01). For hospital admissions, however, the effect remained significant (−0.06, 95% CI −0.11 to −0.03) (table 3, online supplemental figures 5–9). Length of stay decreased numerically as well (−0.24, 95% CI −0.71 to 0.04). Patients seldom experienced IBD-treatment related complications (−0.02, 95% CI −0.05 to 0.01) and long-term corticosteroid use (−0.02, 95% CI −0.04 to 0.01) remained low and unchanged.

Table 3. Healthcare utilisation and ICC by hospital type preimplementation and postimplementation.

	Missing data, %	Hospital type	Period	Mean (95% CI)	ICC, % (p value)	DiD estimate (95% CI)
Healthcare utilisation
ER visits	0	Control	Preimplementation	0.15 (0.10 to 0.20)
			Postimplementation	0.20 (0.13 to 0.28)
		Intervention	Preimplementation	0.21 (0.16 to 0.27)	0.00
			Postimplementation	0.15 (0.11 to 0.19)	0.41 (1)	−0.05 (−0.10 to 0.01)
Admissions	0	Control	Preimplementation	0.12 (0.07 to 0.16)
			Postimplementation	0.16 (0.10 to 0.22)
		Intervention	Preimplementation	0.18 (0.14 to 0.23)	0.00
			Postimplementation	0.12 (0.09 to 0.15)	0.39 (1)	−0.06 (−0.11 to −0.03)
Length of stay, days	0	Control	Preimplementation	11.1 (7.97 to 14.5)
			Postimplementation	8.93 (5.96 to 13.1)
		Intervention	Preimplementation	10.4 (8.62 to 12.2)	0.48
			Postimplementation	7.38 (5.57 to 9.32)	0.81 (1)	−0.24 (−0.71 to 0.04)
Complications	0	Control	Preimplementation	0.10 (0.07 to 0.14)
			Postimplementation	0.07 (0.04 to 0.10)
		Intervention	Preimplementation	0.11 (0.09 to 0.14)	0.00
			Postimplementation	0.05 (0.04 to 0.07)	0.24 (1)	−0.02 (−0.05 to 0.01)
Corticosteroid use (>3 months)	0	Control	Preimplementation	0.05 (0.03 to 0.08)
			Postimplementation	0.03 (0.01 to 0.05)
		Intervention	Preimplementation	0.08 (0.06 to 0.09)	3.63
			Postimplementation	0.04 (0.02 to 0.05)	5.72 (1)	−0.02 (−0.04 to 0.01)

^*For patients who had admissions.

DiD, difference-in-differences; ER, emergency room; ICC, intraclass correlation coefficient.

Symptoms, function, and quality of life

No significant changes were observed for symptoms, function and quality of life following the implementation of the ACP (table 4 and online supplemental figures 10–18). BMI, used as proxy for nutritional status (0.03, 95% CI −0.62 to 0.94), the proportion of patients with an active fistula (0.01, 95% CI −0.05 to 0.02) or anaemia (0.00, 95% CI −0.02 to 0.10) remained stable. Generic quality of life remained stable (0.02, 95% CI −0.14 to 0.05) (table 4).

Table 4. Patient outcomes and ICC by hospital type preimplementation and postimplementation.

	Missing data, %	Hospital type	Period	Mean (95% CI)	ICC, % (p value)	DiD estimate (95% CI)
Symptoms and function
Anaemia	13	Control	Preimplementation	0.16 (0.13 to 0.21)
			Postimplementation	0.15 (0.12 to 0.20)
		Intervention	Preimplementation	0.17 (0.15 to 0.23)	0.00
			Postimplementation	0.17 (0.14 to 0.22)	0.16 (1)	0.00 (−0.02 to 0.10)
BMI, kg/m2	38	Control	Preimplementation	26.6 (24.9 to 29.3)
			Postimplementation	26.8 (29.9 to 30.5)
		Intervention	Preimplementation	26.3 (24.4 to 29.3)	0.31
			Postimplementation	26.5 (24.4 to 31.0)	0.16 (1)	0.03 (−0.62 to 0.94)
Active fistula*	36	Control	Preimplementation	0.13 (0.07 to 0.27)
			Postimplementation	0.09 (0.06 to 0.22)
		Intervention	Preimplementation	0.16 (0.10 to 0.31)	0.00
			Postimplementation	0.13 (0.09 to 0.27)	0.16 (1)	0.01 (−0.05 to 0.02)
Generic quality of life
General health (PROMIS-10)	35	Control	Preimplementation	2.63 (2.33 to 2.98)
			Postimplementation	2.64 (2.41 to 2.88)
		Intervention	Preimplementation	2.66 (2.33 to 3.09)	0.65
			Postimplementation	2.63 (2.35 to 2.97)	0.47 (1)	0.02 (−0.14 to 0.05)
Social health (PROMIS-10)	42	Control	Preimplementation	3.01 (2.47 to 3.53)
			Postimplementation	3.04 (2.51 to 3.58)
		Intervention	Preimplementation	3.03 (2.50 to 3.59)	0.43
			Postimplementation	3.06 (2.46 to 3.64)	0.35 (1)	−0.03 (−0.22 to 0.28)
Physical health (PROMIS-10)	51	Control	Preimplementation	45.0 (38.4 to 51.0)
			Postimplementation	45.1 (34.5 to 55.1)
		Intervention	Preimplementation	44.5 (38.1 to 51.0)	0.76
			Postimplementation	44.8 (34.4 to 54.7)	1.35 (1)	−0.15 (−1.73 to 1.71)
Mental health (PROMIS-10)	44	Control	Preimplementation	45.0 (39.0 to 50.8)
			Postimplementation	45.8 (39.4 to 53.4)
		Intervention	Preimplementation	45.6 (39.5 to 51.7)	0.44
			Postimplementation	46.2 (39.3 to 54.5)	0.50 (1)	0.28 (−1.41 to 1.95)
Utility (EQ-5D-5L)	36	Control	Preimplementation	0.82 (0.72 to 0.87)
			Postimplementation	0.84 (0.76 to 0.88)
		Intervention	Preimplementation	0.82 (0.71 to 0.87)	0.20
			Postimplementation	0.84 (0.74 to 0.88)	0.26 (1)	0.01 (−0.03 to 0.05)
VAS (EQ-5D-5L)	36	Control	Preimplementation	72.8 (64.8 to 77.6)
			Postimplementation	73.2 (65.5 to 77.2)
		Intervention	Preimplementation	71.8 (62.7 to 77.3)	0.48
			Postimplementation	72.6 (63.2 to 77.5)	0.58 (1)	1.28 (−3.64 to 3.93)

^*Only in patients with Crohn’s disease.

BMI, body mass index; DiD, difference-in-differences; EQ-5D-5L, EuroQol Five Dimensions Five Levels questionnaire; ICC, intraclass correlation coefficient; PROMIS-10, 10-item Patient-Reported Outcomes Measurement Information System Global Health Questionnaire; VAS, visual analogue scale.

Variation

Between-hospital variation was minimal at baseline (ICC = ~0%) and remained low postimplementation (ICC = ~1%), indicating that differences in outcomes were mainly driven by unmeasured patient-level related factors rather than hospital-level related factors. The ACP had no effect on the variation between hospitals for the outcomes measured in this study (tables 2–4).

Discussion

An ACP for the treatment of patients with IBD using advanced therapies was implemented to improve IBD care in the southwest region of the Netherlands. This study aimed to assess the effect of an ACP on patient outcomes and healthcare utilisation and to evaluate whether treatment variation changed due to its implementation. Although the statistically significant difference was sustained only for hospital admissions, our findings suggest a potential clinically meaningful effect of the ACP on healthcare utilisation. After 6 months of using the ACP, ER visits, number of admissions and length of hospital stay decreased significantly, all of which are outcomes that are directly relevant to patients with IBD and are quality metrics for IBD care.⁴⁸

Our findings are supported by the results of previous studies in the USA and Europe, which found similar trends or significant reductions in hospitalisation and unplanned care after implementing care pathways or structured IBD services.^{19 49 50} The results point to a consistent pattern across healthcare systems; care pathways may contribute to more efficient and effective IBD management, and healthcare utilisation could be decreased through structured care approaches, regardless of variation in national guidelines or local healthcare infrastructure. However, this must be interpreted with caution. Previous studies had small sample sizes, were observational studies or focused on specific aspects of IBD treatment. Our study was, to our knowledge, the first multicentre prospective quasi-experimental study that was able to evaluate the causal effect of standardising care. However, our sample size calculation was based on the IBD-control questionnaire, which makes the evaluation on healthcare utilisation only explorative. One of the reasons for not finding statistically significant changes in our primary outcome might be due to the methodology of this study. Patients could be included at any point in their treatment trajectory, rather than a standardised clinical milestone such as therapy initiation. As a result, there was no uniform point at which major changes in clinical status could be expected, which likely attenuated observable effects of the ACP on patient outcomes.

With our ACP, we aimed to define a minimum standard of IBD care, based on evidence-based practice where available, without compromising patient outcomes and increasing workload for healthcare providers. While other care models have incorporated additional interventions such as direct-access telephone lines, 24-hour support, patient education programmes or specialty medical homes, our results suggest that implementing a structured pathway potentially already leads to improved outcomes.^{19 49 50} This underscores the importance of first implementing and embedding a robust pathway before layering on resource-intensive interventions, especially considering the current strain on healthcare systems.

Prior literature demonstrated that unwarranted treatment variation is prevalent across IBD centres.^{12–14 51 52} This contrasts with our study; we found no variation between hospitals, likely reflecting regional collaboration, shared training and methodological differences in data analysis.⁹ Due to the already small treatment variation between the hospitals in our study at baseline, we cannot conclude whether standardising care has an effect on the variation. A clinical care pathway for inguinal hernia repair did find significant differences in variation after the implementation, but their baseline variation was considerably higher (~40%–100%), and for the outcomes that had already a low variation, a difference was not observed.¹⁵

This study’s main strengths include its prospective design, the extensive array of outcome measures, a large sample size and the application of robust statistical techniques along with case-mix adjustments to evaluate the impact of the ACP and variation in IBD care across hospitals. Nonetheless, there are limitations to acknowledge. First, the ACP encompassed various protocols for induction of advanced therapies, management of exacerbations and guidelines for annual outpatient visits, but was evaluated as a single entity over 1 year, which may have diluted the impact of individual components. Additionally, even with thorough case-mix adjustments, residual confounding might obscure the actual effect of the ACP and the variation between hospitals.⁵³ Third, the control group might not have been a ‘pure’ control group. Given the regular interdisciplinary meetings between the intervention and control hospitals, and the ACP being based on evidence where available and (inter)national guidelines, it is plausible that elements of the ACP were also adopted by healthcare providers treating patients in the control group. Such contamination could reduce the contrast between groups, potentially attenuating the observed effect of the ACP. Adherence to the ACP and the economic evaluation are assessed in separate articles.^{21 54} At last, the study included only hospitals from the southwest region of the Netherlands; therefore, results cannot be extrapolated directly to other regions of the country. However, as the Dutch healthcare system is uniformly organised nationwide and operates under nationally standardised guidelines, the organisation of IBD care is unlikely to differ substantially between regions. This suggests that our findings may be applicable to other regions of the Netherlands.

Conclusion

Implementation of an ACP for IBD may reduce healthcare utilisation and improve patient-reported remission. Our ACP defined a minimum of care that can achieve outcomes comparable to standard practice, highlighting the value of prioritising structured care pathways, especially given the current burden of IBD on healthcare systems. An ACP can guide healthcare providers and policymakers in standardising treatment and improving quality of care in chronic disease management.

The funders had no role in the design of the study, collection, analysis or interpretation of the data, nor writing of the report.

Data availability statement

Data are available upon reasonable request.