Inflammatory bowel disease outcomes among hospitalized homeless individuals Authors: Tushar Mungle, Vivek Charu, Philip N Okafor Published: 2026 ABSTRACT AbstractBackgroundThere is a paucity of outcome data among homeless individuals with inflammatory bowel diseases (IBD). We reported rates of homelessness among inpatients with Crohn’s disease (CD) and ulcerative colitis (UC) and examined inpatient outcomes.MethodsThe Healthcare Utilization Project (HCUP) State Inpatient Databases from New York, Arizona, Massachusetts and Florida for 2013 and 2014 were used to identify adults ≥ 18 years admitted with IBD identified by ICD-9 codes. Homeless patients were propensity score matched using a one-to-ten greedy nearest-neighbor approach to non-homeless patients to balance the distribution of baseline covariates. Differences in mortality, endoscopy utilization, surgical treatment and post-operative complications were reported as odds ratios with 95% confidence intervals (CI).ResultsA total of 115,008 IBD hospitalizations were identified of which 61% (n = 70,457) were CD discharges while 39% (n = 44,551) were UC discharges. Homelessness was associated with 0.7% (n = 469) of CD discharges and 0.9% (n = 415) of UC discharges. Mortality and endoscopy utilization rates were comparable between homeless and non-homeless patients with CD and UC. Surgical treatment was also broadly similar except for lower odds of bowel resection among homeless IBD patients with UC compared to domiciled IBD patients with UC (OR: 0.41, 95% CI: 0.20 to 0.84). Post-operative complications were also comparable except for lower rates of venous thromboembolism among homeless patients with CD (OR: 0.61, 95% CI: 0.38 to 0.97) whereas it was higher for homeless patients with UC (OR: 1.58, 95% CI: 1.08 to 2.30).ConclusionsHomeless IBD patients have comparable inpatient outcomes with non-homeless IBD patients.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12876-026-04971-7.Keywords: Crohn’s disease, Ulcerative colitis, Endoscopy, Healthcare disparities, Medicaid Article summary Homelessness is associated with 0.7% of Crohn’s disease hospitalizations and 0.9% of ulcerative colitis (UC) hospitalizations. Homeless patients have comparable mortality and endoscopy utilization rates. Surgical outcomes were similar except for lower bowel resection rates among homeless patients with UC. Introduction Homelessness is defined as not having a fixed, regular, or adequate place of residence at nighttime and can be categorized as chronic, intermittent, and transitional depending on the duration [1, 2]. It is an increasingly recognized public health problem, and in 2019 alone, there were at least 568,000 homeless persons on any night in the United States (US) [3]. The interaction between homelessness and health is complex. Homeless individuals experience barriers accessing the health care system and are more likely not to have a primary care physician and less likely to receive preventative health counselling [4]. In addition, when in the healthcare system, homelessness individuals have also reported discrimination and may experience financial and environmental challenges coping with their medical needs [5]. Unsurprisingly, homelessness has been linked with negative health outcomes including higher hospitalization rates, longer length of hospital stay, and higher mortality [6–8]. Healthcare resource utilization has also been shown to be different among homeless individuals [9]. Homeless patients are often uninsured, and the lower utilization of ambulatory health services leads to a higher burden of untreated medical conditions that may require hospitalization [10]. Delays in seeking care medical also leads to an increased use of acute care services [11]. At the same time, about 3.1 million people in the US have Crohn’s disease (CD) and ulcerative colitis (UC) and the prevalence of these inflammatory bowel diseases (IBD) are on the rise [12–15]. While very little is known about the prevalence of IBD among homeless patients, it is established that homeless patients in general are less likely to be seen in clinic where most treatment decisions are made [9], as such, it is conceivable that they may go undiagnosed or untreated when they develop IBD. Presently, there is little to no data on the inpatient management of homeless patients with IBD. It is also unknown how social factors such as housing status impact the decision to hospitalize patients with IBD when they present to the emergency room. There is also a paucity of data on rates of surgical treatment for IBD among patients who are homeless. As a result, in this study, we have utilized a population level dataset to identify hospitalizations for homeless patients with IBD then and compared healthcare resource utilization and inpatient outcomes vs. housed patients with IBD. Describing the rates of IBD in homeless individuals and reporting disparities in health resource utilization is the first step towards achieving health equity for this at-risk group. Materials and methods Data The State Inpatients Databases (SID) are part of a family of databases maintained by the HCUP and sponsored by the Agency for Healthcare Research and Quality (AHRQ) [16]. The study included state inpatient data from Arizona, Florida, New York and Massachusetts for the years 2013 and 2014. Each SID includes all inpatient discharges from non-federal acute-care hospitals and provides information on patient demographics, clinical comorbidities, diagnoses, and expected payer. The data included principal and secondary diagnoses and procedures, health insurance, discharge status, charges, length of stay, patient demographics including sex, age, and race. These states represent different geographic regions of the US and have large homeless populations. Importantly, the homeless population in Massachusetts, Florida and New York is estimated to account for a quarter of the total US homeless population [3]. Since these datasets are de-identified, an IRB exemption was obtained. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnostic codes were used to identify two types of IBD – CD and UC [17]. Further, ICD-9-CM codes and classification of UC and CD subtypes used in analysis are described in Table S1. Inclusion and exclusion criteria Any individual ≥ 18 years with a primary or secondary discharge diagnosis of CD or UC seen between 2013 and 2014 were included in the study. Individuals with both diagnosis codes for Crohn’s disease and ulcerative colitis were excluded from the final cohort to minimize misclassification of IBD subtype. Patients with missing data for age, death, sex, homelessness and race were also excluded from the analysis. Homelessness was identified based on codification of domiciliary status at the time of discharge or death. This determination was made by the hospital of admission and reported to the relevant state and subsequently captured in the SID. Exposure of interest The exposure of interest was homelessness. Each SID discharge record from the states selected includes a designation regarding whether the patient was homeless or not. Demographics and comorbid conditions Patient-level variables included in the study were demographics (age, sex, race, primary payer). Additionally, comorbidities – diabetes, hypertension, chronic kidney disease, carotid artery disease, chronic obstructive pulmonary disease (COPD), asthma, obesity, hyperlipidemia, coronary artery disease, prior myocardial infarction, cerebrovascular disease, prior stroke, stroke, heart failure, atrial fibrillation, flutter, cardiogenic shock, sepsis, acute kidney injury, hemodialysis for acute renal failure (ARF), upper gastrointestinal (GI) bleed, lower GI bleed, hypovolemia, electrolyte imbalance, anemia, malnutrition, total parenteral nutrition, clostridium difficile, blood transfusion, bowel obstruction, chronic steroid use, smoking, pneumonia, mechanical ventilation, cardiac arrest, pulmonary embolism, ischemic bowel, sickle cell disease, human immunodeficiency virus/acquired immunodeficiency syndrome (HIV/AIDS) was also included (Table 1). ICD codes used to identify the relevant comorbid conditions are provided in Table S2. Table 1Baseline characteristics of Crohn's disease homeless and non-homeless cohort, before and after propensity score matchingUnmatchedMatched^Non-homelessHomelessStandardized mean differenceNon-homelessHomelessStandardized mean differencen69,9884694410441Age (%)0.475*0.026 < 4023,068 (33.0)197 (42.0)1870 (42.4)191 (43.3) > = 40 and < 6527,645 (39.5)227 (48.4)2082 (47.2)205 (46.5) > = 65 and < 7510,081 (14.4)24 (5.1)231 (5.2)24 (5.4) > = 759194 (13.1)21 (4.5)227 (5.1)21 (4.8)Female (%)39,851 (56.9)226 (48.2)0.176*2151 (48.8)211 (47.8)0.019 Race (%)0.93*0.04 White55,053 (78.7)175 (37.3)1714 (38.9)175 (39.7) Black6185 (8.8)149 (31.8)1312 (29.8)132 (29.9) Hispanic5702 (8.1)91 (19.4)856 (19.4)84 (19.0) Asian or Pacific Islander501 (0.7)12 (2.6)103 (2.3)8 (1.8) Native American159 (0.2)2 (0.4)19 (0.4)2 (0.5) Other2388 (3.4)40 (8.5)406 (9.2)40 (9.1)Pay1.307*0.058 Medicare28,220 (40.3)120 (25.6)1168 (26.5)119 (27.0) Medicaid11,113 (15.9)296 (63.1)2633 (59.7)269 (61.0) Private insurance24,650 (35.2)13 (2.8)162 (3.7)13 (2.9) Self-Pay3225 (4.6)23 (4.9)267 (6.1)23 (5.2) No charge711 (1.0)9 (1.9)95 (2.2)9 (2.0) Other2069 (3.0)8 (1.7)85 (1.9)8 (1.8)Hospital (%)0.934*0.026 Arizona7077 (10.1)9 (1.9)75 (1.7)9 (2.0) Florida28,932 (41.3)53 (11.3)520 (11.8)53 (12.0) Massachusetts10,106 (14.4)73 (15.6)736 (16.7)73 (16.6) New York23,873 (34.1)334 (71.2)3079 (69.8)306 (69.4)Diabetes (%)9158 (13.1)62 (13.2)0.004550 (12.5)57 (12.9)0.014Hypertension (%)19,834 (28.3)119 (25.4)0.0671104 (25.0)112 (25.4)0.008Chronic kidney disease (%)6878 (9.8)19 (4.1)0.229*196 (4.4)19 (4.3)0.007Carotid artery disease (%)349 (0.5)1 (0.2)0.0487 (0.2)1 (0.2)0.016COPD (%)2229 (3.2)14 (3.0)0.012122 (2.8)13 (2.9)0.011Asthma (%)6398 (9.1)68 (14.5)0.167*643 (14.6)62 (14.1)0.015Obesity (%)5748 (8.2)27 (5.8)0.096257 (5.8)27 (6.1)0.012Hyperlipidemia (%)12,422 (17.7)41 (8.7)0.268*431 (9.8)40 (9.1)0.024Coronary artery disease (%)7579 (10.8)19 (4.1)0.26*192 (4.4)19 (4.3)0.002Prior myocardial infraction (%)2168 (3.1)7 (1.5)0.107*55 (1.2)7 (1.6)0.029Cerebrovascular disease (%)367 (0.5)1 (0.2)0.0517 (0.2)1 (0.2)0.016Prior Stroke (%)3175 (4.5)9 (1.9)0.149*79 (1.8)9 (2.0)0.018Stroke (%)893 (1.3)6 (1.3)< 0.00156 (1.3)5 (1.1)0.012Heart Failure (%)4322 (6.2)17 (3.6)0.118*167 (3.8)16 (3.6)0.008Atrial fibrillation flutter (%)5489 (7.8)9 (1.9)0.278*96 (2.2)9 (2.0)0.009Sepsis (%)5040 (7.2)33 (7.0)0.006305 (6.9)32 (7.3)0.013Acute Kidney Injury (%)7267 (10.4)35 (7.5)0.103*335 (7.6)35 (7.9)0.013Upper GI Bleed (%)3605 (5.2)24 (5.1)0.002211 (4.8)22 (5.0)0.009Lower GI Bleed (%)2484 (3.5)20 (4.3)0.037190 (4.3)18 (4.1)0.011Hypovolemia (%)7403 (10.6)30 (6.4)0.15*301 (6.8)30 (6.8)0.001Electrolyte imbalance (%)14,904 (21.3)67 (14.3)0.184*658 (14.9)66 (15.0)0.001Anemia (%)17,703 (25.3)106 (22.6)0.0631041 (23.6)98 (22.2)0.033Malnutrition (%)5481 (7.8)25 (5.3)0.101*261 (5.9)24 (5.4)0.021Clostridium difficile (%)2304 (3.3)11 (2.3)0.057116 (2.6)11 (2.5)0.009Bowel obstruction (%)6050 (8.6)29 (6.2)0.094267 (6.1)26 (5.9)0.007Chronic steroid use (%)3566 (5.1)5 (1.1)0.235*40 (0.9)5 (1.1)0.023Smoking (%)21,750 (31.1)120 (25.6)0.122*1184 (26.8)117 (26.5)0.007Pneumonia (%)3749 (5.4)9 (1.9)0.184*79 (1.8)9 (2.0)0.018Mechanical ventilation (%)10,036 (14.3)54 (11.5)0.084549 (12.4)52 (11.8)0.02Cardiac Arrest (%)196 (0.3)2 (0.4)0.02519 (0.4)2 (0.5)0.003Pulmonary Embolism (%)595 (0.9)4 (0.9)< 0.00144 (1.0)4 (0.9)0.009Ischemic bowel (%)506 (0.7)2 (0.4)0.03926 (0.6)2 (0.5)0.019HIV/AIDS (%)174 (0.2)10 (2.1)0.174*56 (1.3)6 (1.4)0.008Cardiogenic shock (%)104 (0.1)0 (0.0)0.0550 (0.0)0 (0.0)< 0.001Sickle cell disease (%)68 (0.1)2 (0.4)0.06414 (0.3)1 (0.2)0.017COPD Chronic obstructive pulmonary disease, GI Gastrointestinal, HIV/AIDS Human immunodeficiency virus/Acquired immunodeficiency syndrome*Covariates with imbalance^1:10 (homeless:non-homeless) matching with 29 observations dropped. After matching (469–28) = 441 Baseline characteristics of Crohn's disease homeless and non-homeless cohort, before and after propensity score matching COPD Chronic obstructive pulmonary disease, GI Gastrointestinal, HIV/AIDS Human immunodeficiency virus/Acquired immunodeficiency syndrome *Covariates with imbalance ^1:10 (homeless:non-homeless) matching with 29 observations dropped. After matching (469–28) = 441 Outcomes of interest The primary outcomes were (1) death, (2) endoscopy (esophagogastroduodenoscopy [EGD], colonoscopy, and flexible sigmoidoscopy) utilization (3) rates of surgical procedures including- bowel resection, rectal resections, fistula repair, perianal abscess drainage, pouch, proctocolectomy and temporary ileostomy, ileorectal anastomosis, total abdominal colectomy, colostomy and ileostomy, and (4) rates of postoperative complications between the homeless and non-homeless groups. The SIDs do not contain outpatient or inpatient medication use history but we were able to identify colorectal surgery procedures consisting of operating on a part or whole colon and/or rectum for which ICD-9-CM procedure codes and its classification are described in Table S3. Postoperative complications were evaluated using ICD diagnostic codes, as described in Table S4. Statistical analysis Categorical variables were summarized with counts and proportions. To balance the distribution of measured baseline covariates between homeless and non-homeless patients, the predicted probability (propensity score) for homeless based on comorbidities as covariates (Table S2) was calculated using a logistic regression model. Using one-to-ten greedy nearest-neighbor matching without replacement method, the estimated propensity scores were used to match each homeless patient with ten non-homeless patients. A maximum acceptable difference of 0.2 (caliper width) in propensity score between matched groups was considered. Although we use the term “propensity score”, we note that there are many variables that were predictive (or causative) of homelessness that are not captured in our dataset, and the propensity score in this context is best understood as a balancing score – with the aim of creating covariate balance among the homeless and domiciled cohorts. To assess the balance of the covariates, standardized mean differences (SMDs) for each covariate were calculated before and after propensity score matching (PSM). A SMD value of 0.1 or less indicates a negligible difference or a well-balanced covariate. Using the propensity score matched cohort, the association of outcome was assessed via odds ratio (OR) and 95% confidence interval (CI) and represented via forest plots. P-values less than 0.05 were considered significant. All the statistical analyses were performed using R (version 4.3.1) and “Matching” package for propensity score matching. Results A total of 124,869 discharge records with IBD diagnosis codes were identified for 2013 and 2014 SID from four states: Arizona, Florida, Massachusetts and New York; 115,008 (92.1%) met the inclusion/exclusion criteria for our study (Fig. 1). We identified 70,457 (61%) discharge records with CD, of which 469 occurred in homeless patients (0.66%), and 44,551 records with UC, of which 415 occurred in homeless patients (0.93%; Fig. 1). Consort diagram Clinical and demographic data derived from inpatient discharge records of homeless and domiciled patients with CD are presented in Table 1. Compared to domiciled patients, homeless patients with CD were younger (< 40 years – homeless vs non-homeless: 42% vs 33%), more likely to be White, more likely to have Medicaid (homeless vs non-homeless: 63.1% vs 15.9%). They are less likely to have diagnostic codes for several medical comorbidities, including hyperlipidemia, coronary artery disease, heart failure, atrial fibrillation, and chronic kidney disease, among others (Table 1). Of all the states included in this study, New York state had the highest proportion of homelessness (71.2%) compared to Massachusetts (15.5%), Florida (11.3%) and Arizona (1.9%) (Table 1). To derive a cohort of domiciled patients with a similar distribution of clinical and demographic variables, we fit a propensity score to model the probability of being homeless, conditional on the clinical and demographic variables in Table 1. Applying propensity score matching (with 10 domiciled controls for every 1 homeless patient), we were able to obtain matches for 441 homeless patients (94%)- with 4410 matched controls. The distribution of clinical and demographic data was similar in the homeless and domiciled cohorts after matching (Table 1, Figure S1). We compared several inpatient outcomes between homeless patients with CD and propensity-score matched controls. Inpatient mortality among homeless and non-homeless patient with CD were comparable. We found no evidence for differential utilization of inpatient endoscopy between domiciled vs homeless patients (EGD: 4.17% vs 3.4%, p = 0.437 and for colonoscopy 1.54% vs 1.36% (p = 0.767) (Fig. 2, Table S5). Bowel resection rates were also lower among homeless patients with Crohn’s disease though this did not meet statistical significance (OR: 0.58, 95% CI: 0.34 to 1.01, p = 0.053) (Table S5). Post-operative complications were also broadly similar between homeless and domiciled patients with CD, except for venous thromboembolism, for which the odds of the complication were lower among homeless patients (OR: 0.61, 95% CI: 0.38 to 0.97, p = 0.038) (Table S5). Forest plot of odds ratios (ORs) with 95% confidence intervals comparing inpatient outcomes among hospitalized patients with Crohn’s disease experiencing homelessness versus non-homeless patients in both unmatched and propensity score–matched cohorts. Outcomes assessed included inpatient mortality, endoscopic procedures (EGD, colonoscopy, sigmoidoscopy), postoperative complications, and surgical complications. ORs greater than 1 indicate higher odds among homeless patients, whereas ORs less than 1 indicate lower odds. In the matched analysis, only venous thromboembolism remained significantly associated with homelessness, while the remaining outcomes were not significantly different between groups Clinical and demographic data derived from inpatient discharge records of homeless and domiciled patients with UC are presented in Table 2. These clinical/demographic differences dissipated in the propensity score matched cohort, where a total of 340/415 (82%) homeless UC patients were matched to 3400 domiciled controls (Table 2, Figure S2). Table 2Baseline characteristics of ulcerative colitis homeless and non-homeless cohort, before and after propensity score matchingUnmatchedMatched^Non-homelessHomelessStandardized mean differenceNon-homelessHomelessStandardized mean differencen44,1364153400340Age (%)0.441*0.079 < 4010,682 (24.2)138 (33.3)1105 (32.5)116 (34.1) > = 40 and < 6515,844 (35.9)192 (46.3)1539 (45.3)141 (41.5) > = 65 and < 757997 (18.1)45 (10.8)398 (11.7)43 (12.6) > = 759613 (21.8)40 (9.6)358 (10.5)40 (11.8)Female (%)24,211 (54.9)205 (49.4)0.109*1729 (50.9)171 (50.3)0.011Race (%)1.112*0.084 White32,970 (74.7)111 (26.7)1076 (31.6)111 (32.6) Black3728 (8.4)140 (33.7)974 (28.6)95 (27.9) Hispanic4754 (10.8)104 (25.1)817 (24.0)88 (25.9) Asian or Pacific Islander556 (1.3)17 (4.1)142 (4.2)10 (2.9) Native American147 (0.3)1 (0.2)9 (0.3)1 (0.3) Other1981 (4.5)42 (10.1)382 (11.2)35 (10.3)Pay1.456*0.088 Medicare19,644 (44.5)109 (26.3)983 (28.9)109 (32.1) Medicaid5449 (12.3)273 (65.8)2031 (59.7)198 (58.2) Private insurance15,625 (35.4)13 (3.1)175 (5.1)13 (3.8) Self-Pay1692 (3.8)13 (3.1)136 (4.0)13 (3.8) No charge360 (0.8)1 (0.2)12 (0.4)1 (0.3) Other1366 (3.1)6 (1.4)63 (1.9)6 (1.8)Hospital (%)1.112*0.056 Arizona4769 (10.8)9 (2.2)78 (2.3)9 (2.6) Florida16,229 (36.8)23 (5.5)261 (7.7)23 (6.8) Massachusetts6238 (14.1)35 (8.4)388 (11.4)35 (10.3) New York16,900 (38.3)348 (83.9)2673 (78.6)273 (80.3) Diabetes (%)7966 (18.0)62 (14.9)0.084590 (17.4)57 (16.8)0.016Hypertension (%)15,105 (34.2)125 (30.1)0.0881079 (31.7)110 (32.4)0.013Chronic kidney disease (%)4521 (10.2)24 (5.8)0.165*213 (6.3)23 (6.8)0.02Carotid artery disease (%)326 (0.7)0 (0.0)0.122*0 (0.0)0 (0.0)< 0.001COPD (%)1420 (3.2)8 (1.9)0.08268 (2.0)7 (2.1)0.004Asthma (%)3715 (8.4)37 (8.9)0.018336 (9.9)33 (9.7)0.006Obesity (%)3772 (8.5)16 (3.9)0.195*155 (4.6)16 (4.7)0.007Hyperlipidemia (%)11,287 (25.6)54 (13.0)0.322*529 (15.6)53 (15.6)0.001Coronary artery disease (%)6639 (15.0)31 (7.5)0.241*309 (9.1)29 (8.5)0.02Prior myocardial infraction (%)1659 (3.8)7 (1.7)0.128*66 (1.9)5 (1.5)0.036Cerebrovascular disease (%)345 (0.8)0 (0.0)0.126*0 (0.0)0 (0.0)< 0.001Prior Stroke (%)2425 (5.5)17 (4.1)0.065156 (4.6)15 (4.4)0.009Stroke (%)815 (1.8)3 (0.7)0.132 (0.9)3 (0.9)0.006Heart Failure (%)3927 (8.9)17 (4.1)0.196*151 (4.4)15 (4.4)0.001Atrial fibrillation flutter (%)5149 (11.7)17 (4.1)0.284*178 (5.2)17 (5.0)0.011Sepsis (%)4403 (10.0)47 (11.3)0.044380 (11.2)40 (11.8)0.018Acute Kidney Injury (%)5696 (12.9)51 (12.3)0.019402 (11.8)46 (13.5)0.051Upper GI Bleed (%)2523 (5.7)37 (8.9)0.123*282 (8.3)30 (8.8)0.019Lower GI Bleed (%)2299 (5.2)29 (7.0)0.074220 (6.5)23 (6.8)0.012Hypovolemia (%)5306 (12.0)42 (10.1)0.061349 (10.3)36 (10.6)0.011Electrolyte imbalance (%)10,828 (24.5)89 (21.4)0.073753 (22.1)80 (23.5)0.033Anemia (%)13,421 (30.4)123 (29.6)0.017987 (29.0)92 (27.1)0.044Malnutrition (%)3666 (8.3)31 (7.5)0.031260 (7.6)25 (7.4)0.011Clostridium difficile (%)3699 (8.4)37 (8.9)0.019294 (8.6)34 (10.0)0.047Bowel obstruction (%)1374 (3.1)8 (1.9)0.07679 (2.3)7 (2.1)0.018Chronic steroid use (%)2308 (5.2)8 (1.9)0.178*81 (2.4)8 (2.4)0.002Smoking (%)10,948 (24.8)98 (23.6)0.028803 (23.6)84 (24.7)0.025Pneumonia (%)2694 (6.1)18 (4.3)0.079154 (4.5)18 (5.3)0.035Mechanical ventilation (%)5315 (12.0)26 (6.3)0.201*251 (7.4)26 (7.6)0.01Cardiac Arrest (%)224 (0.5)2 (0.5)0.00413 (0.4)2 (0.6)0.03Pulmonary Embolism (%)556 (1.3)6 (1.4)0.01645 (1.3)6 (1.8)0.036Ischemic bowel (%)664 (1.5)6 (1.4)0.00546 (1.4)6 (1.8)0.033HIV/AIDS (%)276 (0.6)12 (2.9)0.173*80 (2.4)5 (1.5)0.064Cardiogenic shock (%)123 (0.3)2 (0.5)0.03315 (0.4)2 (0.6)0.021Sickle cell disease (%)88 (0.2)0 (0.0)0.0630 (0.0)0 (0.0)< 0.001COPD Chronic obstructive pulmonary disease, GI Gastrointestinal, HIV/AIDS Human immunodeficiency virus/Acquired immunodeficiency syndrome*Covariates with imbalance^1:10 matching with 75 observations dropped. After matching (415–75) = 340 Baseline characteristics of ulcerative colitis homeless and non-homeless cohort, before and after propensity score matching COPD Chronic obstructive pulmonary disease, GI Gastrointestinal, HIV/AIDS Human immunodeficiency virus/Acquired immunodeficiency syndrome *Covariates with imbalance ^1:10 matching with 75 observations dropped. After matching (415–75) = 340 Among patients with UC, there were no significant differences in inpatient mortality and utilization of inpatient endoscopy between homeless and domiciled patients (2.05% vs 2.88%, p = 0.383) (Fig. 3, Table S6). While post-operative complications were broadly similar between the two groups, there was a slightly higher odds for venous thromboembolism among homeless UC patients compared to domiciled controls (OR: 1.58, 95% CI: 1.08 to 2.30, p = 0.019; Fig. 3, Table S6). Considering surgical outcomes, we found evidence that the odds of bowel resection was lower in homeless than in domiciled patients (homeless vs non-homeless: 2.35% vs 5.53%; OR: 0.41; 95% CI: 0.20 to 0.84, p = 0.015) (Table S6), but the remainder of surgical outcomes considered were comparable between the two groups. Forest plot of odds ratios (ORs) with 95% confidence intervals comparing inpatient outcomes among hospitalized patients with Ulcerative colitis experiencing homelessness versus non-homeless patients in both unmatched and propensity score–matched cohorts. Outcomes assessed included inpatient mortality, endoscopic procedures (EGD, colonoscopy, sigmoidoscopy), postoperative complications, and surgical complications. ORs greater than 1 indicate higher odds among homeless patients, whereas ORs less than 1 indicate lower odds. In the matched analysis, bowel resection and venous thromboembolism remained significantly associated with homelessness, while the remaining outcomes were not significantly different between groups Discussion In this multistate analysis using administrative claims data, we compared inpatient outcomes between homeless vs. domiciled individuals with IBD including CD and UC in the states of Arizona, Florida, Massachusetts, and New York. Using a propensity matching approach to account for demographic and clinical differences in both populations, we observed mostly comparable outcomes between homeless vs. non-homeless individuals. In particular, we observed similar odds of inpatient death and endoscopy utilization between homeless patients with IBD and domiciled controls. While overall inpatient surgical utilization was similar between the two groups, we did find evidence that the odds of bowel resection was lower among homeless patients with UC compared to domiciled controls. Studies have shown that homeless individuals may have as much as eight comorbidities at any time [18] with the most prevalent health issues being respiratory, dermal, injuries, and digestive problems [19]. To our knowledge, this is the among the first large multistate administrative database studies comparing inpatient outcomes between homeless and domiciled patients hospitalized with IBD. This vulnerable patient group is understudied likely because of the dearth of databases that capture the housing status of patients, especially at a population level, and the variation in how homelessness is defined. In addition, we found no prior studies on medication prescription patterns among homeless patients with IBD. The few studies that described prescription patterns among homeless individuals did not report IBD medication utilization [20]. Given the nature of the SID, which do not capture medication data, we could only reliably explore endoscopy and surgical utilization rates. Surgical treatment for patients with IBD is often a last resort when medical therapies have failed or in clinical scenarios where there are complications due to disease progression. We compared rates of procedures that typically require inpatient stay including bowel resection, perianal abscess drainage, colectomy, colostomy, and ileostomy. Interestingly, we found significantly lower rates of bowel resection among homeless patients with UC. Additionally, rates of bowel resection among homeless patients with CD were also lower but this did not meet statistical significance. No difference was seen in colostomy and ileostomy rates, and among patients with CD or UC, no difference was seen in rates of fistulae repair and perianal abscess drainage between homeless and non-homeless patients. The reasons for the lower rates of bowel resection among the homeless patients with UC remain unclear. One possible explanation could be that homeless patients with UC may present with less severe disease on admission. The absence of medication and endoscopy data in the SID limited our ability to stratify by disease severity. Given the nature of the SID, we were also unable to assess the impact of homelessness and other socioeconomic determinants of health on the decision by providers to admit these patients for reasons that are more social than medical. It is possible that the observed differences in surgical resection rates stem from a provider or health system perspective due to unmeasured confounders such as implicit bias and stigmatization. In a systematic review on disparities in surgery for IBD, Booth et al., showed that individuals from socioeconomically disadvantaged backgrounds where less likely to undergo abdominal surgery when hospitalized [21]. They concluded that surgical disparities based on sociological and structural factors may suggest unidentified differences in multidisciplinary care. Lesperance et al., in a population level analysis showed that insurance status and hospital teaching status were significant predictors of receiving minimally invasive approaches to surgery [22]. The impact of health insurance on the receipt of bowel resection has also been demonstrated using population level data [23]. Studies on insurance coverage among homeless have shown that up to 50% may have Medicaid as the primary payer, while rates of self-pay can be as high as 3% [24]. These rates are comparable with our findings. Unfortunately, the state inpatient databases used for this study did not provide hospital teaching status to allow us explore the impact of hospital characteristics such as size and teaching status on the receipt of bowel resection among the subpopulation of patients with IBD who are homeless. We also could not explore the impact of urban or rural dwelling on outcomes among the homeless cohort. We also observed similar rates of post-operative complications between homeless and non-homeless patients with IBD except for rates of venous thromboembolism (VTE) which were higher among homeless patients with ulcerative colitis, but lower in the CD cohort. This observation is worthy of discussion. The mechanism behind higher rates of VTE among patients hospitalized with IBD has not been fully elucidated but various risk factors have been implicated especially disease activity which increases systemic inflammation [25]. Our observation could be explained by lower inflammatory burden among CD patients undergoing surgery vs. those with UC undergoing surgery who have typically failed to respond to inpatient medical management. Another explanation could be differential rates of venous thromboembolism prophylaxis in homeless patients with ulcerative colitis especially if they present with bloody diarrhea. Again, the absence of medications, laboratory and imaging data within the state inpatient databases does not allow further exploration of these hypotheses. However, other studies have also demonstrated higher rates of VTE among hospitalized patients with ulcerative colitis compared to patients with Crohn’s disease including after bowel resection [26, 27]. Importantly, the cited studies did not look at the impact of homelessness on these outcomes. Among studies that have looked at health outcomes among homeless patients, many have shown worse outcomes compared to domiciled patients. Wadhera et al. reported higher mortality rates among homeless individuals admitted for acute cardiovascular diseases and also showed lower rates of treatment with potentially life-saving procedures including percutaneous coronary interventions and coronary artery bypass grafting among homeless patients which they alluded could explain their observed mortality differences [28]. Higher mortality rates have also been reported in homeless patients hospitalized with acute gastrointestinal bleeding [24]. We found comparable endoscopy rates (< 5% for both upper endoscopy and colonoscopy) in our homeless and non-homeless cohorts with either CD or ulcerative colitis. Similar studies among homeless patients hospitalized for gastrointestinal bleeding also did not show any significant difference in endoscopy resource utilization when compared to patients who were not homeless at the time of admission [24]. Studies have shown that homeless patients can be refused health care and filling of prescriptions even when they have primary coverage because of lack of identifying information, lack of insurance, or the burden of prescription costs that fall on the patient even with insurance coverage [7]. In addition, homeless patients often get their medical care at emergency rooms as opposed to primacy care provider offices. We were unable to study emergency room utilization given the nature of the state inpatient databases. With complex medical illnesses such as IBD, most of the longitudinal care is managed by specialist gastroenterologists. We were unable to find any studies that looked at specialist GI care utilization among homeless patients with inflammatory bowel diseases. However, among patients with other chronic illnesses such as diabetes and hypertension, rates of poorly controlled disease are higher among homeless patients in the US [29]. This is supported by studies among patients with other chronic medical conditions like diabetes, that have shown that homeless people expressed higher rates of difficulty in managing their medical conditions [30]. The major limitation of our study is the sample size of homeless patients with an IBD diagnosis and the solely inpatient nature of the databases. Although propensity score matching improved covariate balance, the matching process reduced the effective sample size. Consequently, our study may have been underpowered to detect modest but clinically meaningful differences for infrequent outcomes such as specific surgical procedures and postoperative complications. Several outcome estimates demonstrated wide confidence intervals, and therefore nonsignificant findings should not be interpreted as evidence of absence of effect. For example, bowel resection among homeless patients with Crohn’s disease demonstrated a lower point estimate (OR 0.58, 95% CI 0.34—1.01) that may still represent a clinically meaningful difference despite narrowly missing conventional statistical significance. This was largely because we were only able to study hospitalizations in a few states because most of the states under the healthcare cost and utilization project do not codify homelessness. Also, among those that coded for homelessness, this was only done for certain years making it difficult to explore outcomes on a longitudinal basis. However, prior research shows that the states included in our analysis may account for a quarter of the US homeless population [28]. There is also a possibility of accuracy of the diagnostic codes used in the SID, while codes for IBD have been shown to be reliable, the reliability of coding for homelessness has under studied. Another limitation to strongly acknowledge may be that of selection bias, given the inpatient nature of the databases utilized for this study, it is possible that patients with more severe disease were overrepresented in the nonhomeless subgroup as opposed to the homeless subgroup which may have a different disease severity mix because of the impact of homelessness. The absence of medication and endoscopy data prevents us from better controlling for this bias. To minimize misclassification bias, we applied restrictive selection criteria during our analysis cohort creation by excluding patients with both UC and CD diagnoses which further reduced our sample size. More research is needed in standardizing how homelessness is identified and coded in large databases. It is also essential to better understand how sociodemographic variables are captured across providers, systems, and states to minimize the possibility of misclassification [31]. We were also unable to ascertain duration of homelessness and IBD, and pharmacotherapeutic history given the nature of the database. The lack of medication data limits the interpretation of our findings as we were unable to assess disease severity and unable to explore undertreatment of patients in the homeless cohort. In addition, we were unable to study the impact of homelessness on outpatient IBD outcomes. We were also unable to investigate geographic and facility level differences in care for homeless patients with IBD given the nature of the state inpatient datasets. We acknowledge the possibility of residual selection bias as it is likely that only patients with severe IBD would have been hospitalized. A subset of homeless patients could not be matched within the predefined propensity score caliper, suggesting baseline differences relative to available non-homeless controls. Exclusion of these unmatched patients may limit generalizability and could introduce residual selection bias if these individuals differed systematically in disease severity, comorbidity burden, or healthcare access. We were unable to adequately control for unmeasured confounders such as delays in access to care, adherence to provider recommendations, stigmatization, and implicit biases. We were also unable to distinguish if homelessness status was permanent or temporary and how care for homeless patients admitted with IBD transitioned to the outpatient setting. Finally, homelessness status in the SID depended on codification by the discharge hospital. As such, it is possible that some patients were incorrectly classified or misclassified according to varying definitions of homelessness across institutions. Despite these limitations, this is the first study reporting health outcomes in homeless patients with IBD at a population level and our results highlight a few important points. Our findings, despite the aforementioned limitations have several points of clinical relevance. First, health services and systems need to improve the identification of vulnerable individuals including the homeless so that outcomes can be better studied at a broader level. Identification of high-risk homeless patients will afford optimization of existing models of care targeted to homeless patients e.g. hiring of GI specific social workers and case managers to help facilitate these patients as they transition the care continuum, leveraging mobile technology to improve IBD clinic utilization, medication compliance and adherence, as the majority of homeless patients in certain states own mobile phones [27]. Second, policy initiatives focused on achieving targets dedicated to homeless individuals are needed at the hospital and state level. Third, while our results are mostly reassuring showing similar utilization of endoscopy resources between both groups, more focused qualitative research is needed to better understand the perspectives of homeless patients with IBD, especially regarding their preference of IBD medications and utilization of GI health resources. For instance, it is essential to understand the impact of route and frequency of biologics and other small molecules on the decision to receive IBD medical therapy especially with the increasing availability of novel oral agents. In conclusion, we have shown that homeless patients with IBD, for the most part, receive inpatient endoscopic and surgical care at rates that are comparable to their non-IBD counterparts. It is important to emphasize that the nature of the databases used for the analysis might have limited our ability to identify true differences between both groups of patients and we encourage further research on this topic. While our sample size is modest because of the intrinsic limitations associated with the codification of homelessness status in existing state inpatient databases, our results form a basis for future outcomes and health policy research. It would otherwise be difficult to study outcomes among homeless patients with IBD as many health systems presently focus more on acute health issues of the homeless vs. chronic disease management [32]. More research is needed on outpatient outcomes including the utilization of GI-specific health resources among homeless patients with IBD, but we acknowledge that this may be difficult to achieve at this time until hospitals and health systems consistently track outcomes of all their vulnerable patients. Supplementary Information Supplementary Material 1: Figure S1. Standardize mean difference illustrating balance of covariates before and after propensity score matching for Crohn’s disease. Supplementary Material 2: Figure S2. Standardize mean difference illustrating balance of covariates before and after propensity score matching for ulcerative colitis. Supplementary Material 3: Supplemental Tables. Acknowledgements None. Abbreviations Inflammatory Bowel Diseases Crohn’s Disease Ulcerative Colitis Confidence Interval Esophagogastroduodenoscopy International Classification of Diseases, Ninth Revision, Clinical Modification Institutional Review Board Odds ratio State inpatient database Authors’ contributions T.M (data analysis and interpretation, manuscript preparation), V.C. (study design, data analysis and interpretation, manuscript preparation), P.O (study design, manuscript preparation). Funding This work was supported by Pfizer Quality Improvement Grant 68494035. Data availability The third-party datasets analyzed for this study are available from the Healthcare Cost and Utilization Project State Inpatient Databases at [HCUP-US SID Overview (ahrq.gov)] (https:/hcup-us.ahrq.gov/sidoverview.jsp). Declarations Ethics approval and consent to participate Since these datasets are de-identified, an IRB exemption was obtained. Consent for publication Not Applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Vivek Charu and Philip N. Okafor are co-senior authors. References Associated Data Supplementary Materials Supplementary Material 1: Figure S1. Standardize mean difference illustrating balance of covariates before and after propensity score matching for Crohn’s disease. Supplementary Material 2: Figure S2. Standardize mean difference illustrating balance of covariates before and after propensity score matching for ulcerative colitis. Supplementary Material 3: Supplemental Tables. Data Availability Statement The third-party datasets analyzed for this study are available from the Healthcare Cost and Utilization Project State Inpatient Databases at [HCUP-US SID Overview (ahrq.gov)] (https:/hcup-us.ahrq.gov/sidoverview.jsp).