Patterns of Multimorbidity among Older Adults with Colorectal Cancer Living in Puerto Rico

Abstract

Background:

Multimorbidity among Hispanic colorectal cancer (CRC) patients in Puerto Rico remain poorly understood despite their potential impact on patient outcomes. We examined patterns of chronic conditions at CRC diagnosis and their associations with sociodemographic factors, stage at diagnosis, and tumor location.

Methods:

We used the Puerto Rico Central Cancer Registry–Health Insurance Linkage Database (2012–2020). Adults aged ≥50 years with a first CRC diagnosis were included. Chronic conditions were identified using Medicare & Medicaid Chronic Conditions Warehouse algorithms. Latent class analysis and regression models assessed multimorbidity patterns and associations with sociodemographic and tumor characteristics.

Results:

Among 6,543 adults with CRC (67.7 years mean age; 45.7% female), 52.7% had no additional chronic conditions, 18.7% had one, and 28.6% had ≥2 conditions. The most prevalent comorbidities were hypertension (20.6%), diabetes (14.4%), and hyperlipidemia (12.2%). Three multimorbidity patterns were identified: Hypertensive Vision–Joint, Diabetic Renal, and Cardiopulmonary Severe. Advanced age, female sex, and Medicare insurance predicted class membership. Older age was associated with membership in the Cardiopulmonary Severe group (aPR=1.75 for ages 76–85 years; aPR=2.86 for ≥86 years). Females were more likely to belong to the Hypertensive Vision–Joint group (aPR=1.42). Multimorbidity patterns were not associated with tumor location or stage at diagnosis.

Conclusions:

Multimorbidity is common among older Hispanic adults with CRC in Puerto Rico and forms distinct patterns associated with sociodemographic factors, highlighting heterogeneity in medical complexity.

Impact:

Identifying multimorbidity patterns among older CRC patients in Puerto Rico underscores the need for integrated, patient-centered care for aging Hispanics.

Introduction

Colorectal cancer (CRC) remains the second leading cause of cancer deaths in the United States, accounting for an estimated 55,230 deaths in 2026.¹ Advances in CRC prevention, screening, and treatment have contributed to improvements in clinical outcomes for patients.² However, profound inequalities in CRC incidence and mortality persist among Hispanic adults.³ In Puerto Rico, where approximately 43.5% of the population lives in poverty, CRC is the second leading cause of cancer death among Hispanic men and women.⁴ Compared to the United states mainland, individuals in Puerto Rico face higher incidence of early-onset CRC, more frequent diagnosis at advanced stages, and worse survival outcomes.^5,6

Older adults carry the greatest burden of CRC morbidity and mortality.⁷ Many also experience multimorbidity, defined as the presence of two or more chronic conditions.⁸ Conditions such as diabetes, hypertension, and hyperlipidemia have been associated with increased CRC risk and mortality.⁹ These conditions may also delay symptom recognition, complicate treatment decision-making, and worsen clinical outcomes.⁹ At the time of diagnosis, multimorbidity can directly affect the course of CRC treatment by limiting eligibility for surgery or systemic therapy, increasing the risk of treatment-related toxicity, necessitating dose modifications or delays, and shifting care toward less aggressive or more palliative approaches.¹⁰ Understanding how chronic conditions are distributed at the time of CRC diagnosis, and how they cluster within individuals, is essential for identifying patients with complex medical needs.¹¹ This knowledge can help inform personalized care plans by aligning treatment intensity with patients’ overall health status, optimizing management of coexisting conditions during cancer therapy, and improving clinical decision-making.¹¹ However, few studies have focused on Hispanic populations or individuals in resource-limited settings.^12–22 Additionally, older adults with multiple chronic conditions are often excluded from clinical trials, limiting the evidence base to guide care in this high-risk group.²³

This study aimed to examine multimorbidity among older adults with CRC in Puerto Rico. Specifically, we (1) estimated the prevalence of chronic conditions at the time of CRC diagnosis, (2) identified common multimorbidity patterns using latent class analysis, (3) evaluated sociodemographic predictors of multimorbidity patterns, and (4) assessed the association of multimorbidity patterns with cancer stage at diagnosis and tumor location.

Materials and Methods

Data source.

In this retrospective cohort study, we used the Puerto Rico Central Cancer Registry-Health Insurance Linkage Database (PRCCR-HILD) from 2012 to 2020. The PRCCR is part of the Centers for Disease Control and Prevention’s National Program of Cancer Registries and follows established coding standards. By law, all healthcare facilities in Puerto Rico must report cancer cases to the PRCCR, which collects clinical and demographic data from hospitals, outpatient clinics, pathology laboratories, and radiotherapy/chemotherapy centers across the island. The PRCCR identifies cases through hospital records, pathology laboratories, and the Vital Statistics Office, where a legally enforced death registration system ensures data completeness. Recent audits confirm that the PRCCR maintains high case ascertainment rates, comparable to the national median. To link PRCCR data with health insurance databases, PRCCR staff use a deterministic matching approach similar to the one used by SEER-Medicare, ensuring comprehensive coverage of the insured population in Puerto Rico, estimated at 92%.²⁴ PRCCR-HILD includes Medicare, Medicaid, and private insurance claims. The Institutional Review Boards of the University of Massachusetts Chan Medical School and the PRCCR approved this study. Because the analysis relied on linked registry and administrative claims data at the time of diagnosis, there was no participant loss to follow-up after cohort definition. Sample reduction occurred only through predefined eligibility exclusions, minimizing the potential for attrition bias. This was an observational study using existing data sources; therefore, no randomization or experimental allocation of participants was performed, and blinding of investigators or participants was not applicable.

Study population.

This analysis included residents of Puerto Rico with a first primary diagnosis of CRC, as defined by the International Classification of Diseases for Oncology, Third Edition (ICD-O-3).²⁵ The study population consisted of adults diagnosed with malignant CRC between January 1, 2012, and December 31, 2020, with successfully matched health insurance claims (n = 7,470). We excluded individuals under 50 years of age (n = 927). While CRC screening is now recommended to begin at age 45 years, our focus on individuals aged 50 years and older reflects the screening guidelines in place during most of the study period. Because the American Cancer Society (ACS) and the U.S. Preventive Services Task Force (USPSTF) updated their recommendations in 2018 and 2021 respectively.^26,27 Individuals younger than 50 years were excluded because this study focused on multimorbidity patterns among older adults, who bear the greatest burden of CRC and chronic disease complexity. In addition, claims-based multimorbidity ascertainment algorithms are more consistently captured in older insured populations. The final analytic sample included 6,543 individuals. The study sample size was determined by the number of eligible CRC cases meeting inclusion criteria during the study period, and no a priori power calculation was conducted.

Operational Definitions of Variables

Chronic Conditions:

We identified 21 non-cancer chronic conditions using diagnosis and procedure codes from the PRCCR-HILD administrative data, applying the Chronic Conditions Warehouse (CCW) algorithms.^28,29 Chronic conditions were ascertained from inpatient, outpatient, and physician/supplier claims using the condition-specific look-back periods recommended by the CCW algorithms (generally 1-3 years prior to CRC diagnosis, depending on the condition). The CCW algorithms were developed by the Centers for Medicare & Medicaid Services and are widely used in healthcare databases to identify chronic conditions, particularly among Medicare beneficiaries.^28,29 The conditions assessed included hypertension, hyperlipidemia, cataract, diabetes, glaucoma, rheumatoid arthritis or osteoarthritis, anemia, acquired hypothyroidism, ischemic heart disease, chronic kidney disease, depression, osteoporosis, asthma, chronic obstructive pulmonary disease, acute myocardial infarction, atrial fibrillation, Alzheimer’s disease, dementia, heart failure, hip or pelvic fracture, and stroke.²⁹

Predictors of multimorbidity:

Sociodemographic characteristics were considered predictors of multimorbidity, as they are known to be associated with the accumulation and distribution of chronic conditions across the life course.³⁰ Accordingly, the following variables measured at the time of CRC diagnosis were included: age at diagnosis (years, mean ± SD), sex at birth, and marital status at diagnosis. Sex was recorded in the PRCCR as sex assigned at birth (male or female) and was included as a biological variable in all analyses. Marital status was categorized as married (including common-law or domestic partnership) or unmarried (never married, separated, divorced, or widowed). Health insurance type was included as a proxy for healthcare access and socioeconomic position and was classified as Medicaid, Medicare, Medicare and Medicaid (dual eligible), or private insurance. Geographic variation in healthcare access and population characteristics was captured using health regions defined by the Puerto Rico Department of Health, including North (Arecibo), Central (Bayamón), Southeast (Caguas), East (Fajardo), West (Mayagüez), Northeast (Metro), and South (Ponce).

Outcomes:

The outcomes of interest were tumor location and stage at diagnosis. Tumor location was classified as colon or rectum, including cancers of the rectosigmoid junction, using International Classification of Diseases for Oncology, Third Edition (ICD-O-3) codes. Stage at diagnosis was defined using two standard classification systems. Clinical stage was classified according to the American Joint Committee on Cancer (AJCC) staging system as stages 0, I, II, III, and IV.³¹ In addition, stage at diagnosis was categorized as localized, regional, or distant based on registry summary stage classifications for analytic purposes.

Statistical analysis.

Descriptive statistics were employed to summarize the sociodemographic and clinical characteristics of the study population. Continuous variables, such as age at diagnosis, were reported as means with standard deviations (SD), while categorical variables were presented as frequencies and percentages. The prevalence of chronic conditions was assessed using CCW algorithms.²⁸ For each approach, the prevalence of individual chronic conditions was calculated. A low disease burden group was created for participants with no or one chronic condition, aligning with the definition of multimorbidity as the presence of two or more chronic conditions.

We performed a latent class analysis (LCA) to identify distinct subgroups among participants with two or more chronic conditions (n=1,873).³² To enhance model stability and interpretability, we limited the analysis to chronic conditions with a prevalence of at least 2% in the study population, as rare conditions provide limited information and can lead to unstable class estimates. This threshold was selected to balance model parsimony with clinical relevance. We determined the optimal number of classes by starting with a one-class model and progressively increasing the number of classes, evaluating model fit using the Akaike Information Criterion (AIC), the Bayesian Information Criterion (BIC), and clinical interpretability.³² The three-class solution provided the best balance between statistical fit and conceptual clarity. Individuals were assigned to the class for which they had the highest posterior probability. No exact ties in posterior probabilities were observed.

Multiple Imputation by Chained Equations (MICE) was used to address missing data in predictor variables, with missingness rates ranging from 0.03% to 18.4%.³³ Twenty imputations were performed to create multiple imputed datasets. MICE is a versatile and effective method for handling missing data, imputing values based on observed data relationships.³³ Analyses were conducted separately within each imputed dataset, and results were combined using Rubin’s rules to obtain adjusted estimates and standard errors.³³ Associations between sociodemographic predictors and multimorbidity patterns (latent class membership) were evaluated using multinomial logistic regression models. Associations between multimorbidity latent classes and cancer-related outcomes were assessed using modified Poisson regression models to report relative prevalence with robust variance estimation. Outcomes for these analyses were dichotomized as advanced versus early stage at diagnosis and colon versus rectal cancer. Results are presented as prevalence ratios and 95% confidence intervals. We assessed non-linearity in age by adding a quadratic term. Non-linearity in the association between age at diagnosis and latent class membership was assessed using a quadratic term and a joint Wald test across outcome categories. Given evidence of non-linearity, age was subsequently modeled as a categorical variable (50–64, 65–75, 76–85, and ≥86 years) for ease of interpretation.

Results

The mean age at diagnosis was 67.7 years (SD ± 9.8), and 45.7% were female. Most patients were married or in a domestic partnership (57.2%), and 35.1% had Medicare. At diagnosis, the majority had colon cancer (72.8%), and 42.6% had localized-stage disease (Table 1).

Table 1.

Clinical and Sociodemographic Characteristics Among Adults Diagnosed with Colorectal Cancer in Puerto Rico, 2012–2020 (n=6,543)

Characteristics	N (%)
Age at diagnosis (years) mean ± SD	67.7 ± 9.8
Female	2,988 (45.7)
Marital status
Single	1,338 (22.6)
Married/Domestic partner	3,384 (57.2)
Separate/Divorced/Widowed	1,192 (20.2)
Insurance at diagnosis
Private	1,137 (17.4)
Medicaid	1,809 (27.8)
Medicare	2,286 (35.1)
Medicare-Medicaid	1,284 (19.7)
Region at diagnosis
Northeast (Metro)	1,290 (19.7)
Central (Bayamon)	1,010 (15.4)
East (Fajardo)	227 (3.5)
North (Arecibo)	876 (13.4)
West (Mayaguez)	1,062 (16.2)
South (Ponce)	1,053 (16.1)
Southeast (Caguas)	1,023 (15.6)
Cancer site
Colon	4,764 (72.8)
Rectum*	1,779 (27.2)
First of multiple cancers	788 (12.0)
Adenocarcinoma	6,389 (97.6)
Clinical stage
0	202 (3.8)
I	1,612 (30.2)
II	1,304 (24.4)
III	1,428 (26.8)
IV	789 (14.8)
Stage at diagnosis
Localized	2,783 (42.6)
Regional	2,920 (44.7)
Distant	826 (12.7)
Grade
Well differentiated	1,111 (28.3)
Moderately differentiated	2,488 (63.5)
Poor differentiated	321 (8.2)

^*Include cancers of the rectosigmoid junction. Missing data: marital status (n=629), insurance (n=27), region (n=2), clinical stage (n=1,208), stage at diagnosis (n=14), grade (n=2,623).

The most common conditions identified were hypertension (20.6%), diabetes (14.4%), and hyperlipidemia (12.2%), and 28.6% of patients with CRC had two or more of those chronic conditions (Figure 1). Model fit indices indicated that the three-class solution provided the best fit (AIC = 24006.7, BIC = 24250.3, Supplemental Table 1). Class 1, labeled Hypertensive Vision-Joint, was characterized by a high prevalence of hypertension (58.3%) and hyperlipidemia (38.3%), along with notable rates of glaucoma, cataracts, and osteoarthritis. Class 2, Diabetic Renal, exhibited the highest prevalence of diabetes (76.9%) and chronic kidney disease (51.7%), reflecting a metabolic-renal disease pattern. Class 3, Cardiopulmonary Severe, showed the highest burden of heart failure (64.0%) and ischemic heart disease (72.3%), along with elevated rates of chronic kidney and pulmonary conditions (Supplemental Table 2).

Figure 1.

Prevalence of Chronic Conditions Among Adults Diagnosed with Colorectal Cancer in Puerto Rico, 2012–2020.

The upper panel displays the distribution of chronic disease burden categorized as 0, 1, or ≥2 chronic conditions among adults diagnosed with colorectal cancer in Puerto Rico (n=6,543). The lower panel displays the prevalence (%) of individual chronic conditions identified using Chronic Conditions Warehouse (CCW) definitions based on linked insurance claims data. Bars represent the proportion of patients with each chronic condition. Rheumatoid arthritis (RA) and osteoarthritis (OA) were combined into a single category. Hip or pelvic fracture cases were excluded from the figure because the prevalence was <11 cases.

Details on sociodemographic and clinical characteristics by latent class group are presented in Supplemental Table 3. Older age at diagnosis was associated with a higher adjusted prevalence of the Hypertensive Vision–Joint and Cardiopulmonary Severe classes, with the strongest associations observed among individuals aged ≥76 years. Women had a higher probability of being classified into the Hypertensive Vision–Joint group compared with men (aPR = 1.42). Regarding insurance status, individuals with Medicare or dual Medicare–Medicaid coverage were more likely to belong to all three latent groups; however, the association was strongest for the Cardiopulmonary Severe group (aPR = 2.93 and 4.37, respectively) compared with those with private insurance. Regional differences were also observed: patients residing in the North and West regions were more likely to be in the Diabetic Renal group (aPR = 1.62 and 1.54, respectively) compared with those from the Central Northeast metropolitan area (Table 2), whereas no associations were observed for the other groups. After adjusting for age, sex, and insurance status, latent class membership showed evidence of association with tumor location. Compared with the reference class, the Hypertensive Vision–Joint class (aPR=0.97, 95% CI 0.95–1.00) and the Diabetic Renal class (IRR=0.95, 95% CI 0.93–0.98) had slightly lower relative rates. No differences were observed for the Cardiopulmonary Severe class. For clinical stage, latent class membership was not associated with advanced stage. For stage at diagnosis, the Diabetic Renal class had a lower relative rate (aPR=0.93, 95% CI 0.86–1.00), while other classes did not differ from the reference group (Supplemental Table 4 and Table 3).

Table 2.

Association of Sociodemographic Characteristics with Latent Class Groups Using Chronic Conditions Warehouse Definitions Among Adults Diagnosed with Colorectal Cancer in Puerto Rico, 2012–2020^*

Characteristics	Crude Prevalence Ratio (95% CI)			Adjusted Prevalence Ratio (95% CI)**
	Hypertensive Vision-Joint	Diabetic Renal	Cardiopulmonary Severe	Hypertensive Vision-Joint	Diabetic Renal	Cardiopulmonary Severe
Age at diagnosis (years)
50-64	1.00	1.00	1.00	1.00	1.00	1.00
65-75	2.32 (1.94-2.78)	2.45 (2.04-2.96)	3.24 (2.12-4.96)	1.35 (1.08-1.68)	1.26 (1.00-1.58)	1.23 (0.75-2.01)
76-85	3.38 (2.76-4.15)	3.11 (2.50-3.86)	4.87 (3.07-7.73)	1.81 (1.41-2.31)	1.53 (1.18-1.98)	1.75 (1.02-2.98)
86+	4.31 (3.15-5.93)	2.70 (1.84-5.93)	8.07 (4.40-14.79)	2.13 (1.49-3.04)	1.34 (0.88-2.03)	2.85 (1.44-5.62)
Sex
Male	1.00	1.00	1.00	1.00	1.00	1.00
Female	1.46 (1.26-1.68)	1.14 (0.98-1.33)	0.89 (0.65-1.21)	1.42 (1.21-1.65)	1.18 (1.00-1.38)	0.82 (0.59-1.14)
Marital status
Single	1.00	1.00	1.00	1.00	1.00	1.00
Married/Domestic partner	1.21 (0.99-1.48)	1.19 (0.97-1.46)	1.19 (0.77-1.84)	1.07 (0.86-1.32)	0.97 (0.78-1.21)	0.92 (0.59-1.44)
Separate/Divorced/Widowed	1.97 (1.57-2.46)	1.49 (1.17-1.89)	1.97 (1.22-3.17)	1.25 (0.98-1.59)	0.94 (0.73-1.22)	1.17 (0.70-1.93)
Insurance at diagnosis
Private	1.00	1.00	1.00	1.00	1.00	1.00
Medicaid	0.62 (0.48-0.81)	0.60 (0.45-0.81)	0.58 (0.26-1.30)	0.59 (0.45-0.77)	0.55 (0.40-0.74)	0.54 (0.24-1.21)
Medicare	2.63 (2.11-3.29)	2.78 (2.18-3.54)	4.63 (2.52-8.54)	1.84 (1.43-2.36)	2.06 (1.57-2.70)	2.93 (1.51-6.68)
Medicare-Medicaid	2.18 (1.70-2.79)	2.93 (2.26-3.80)	6.70 (3.60-12.47)	1.55 (1.18-2.03)	2.10 (1.58-2.81)	4.37 (2.23-8.57)
Region at diagnosis
Northeast (Metro)	1.00	1.00	1.00	1.00	1.00	1.00
Central (Bayamon)	1.17 (0.93-1.48)	1.14 (0.87-1.49)	1.20 (0.72-2.01)	1.25 (0.98-1.60)	1.19 (0.90-1.58)	1.21 (0.72-2.04)
East (Fajardo)	0.93 (0.61-1.42)	1.35 (0.88-2.07)	1.09 (0.45-2.66)	0.93 (0.60-1.45)	1.29 (0.83-2.00)	0.97 (0.39-2.38)
North (Arecibo)	0.77 (0.59-1.01)	1.53 (1.18-1.99)	0.75 (0.41-1.37)	0.85 (0.64-1.12)	1.62 (1.23-2.13)	0.72 (0.39-1.34)
West (Mayaguez)	0.93 (0.73-1.19)	1.50 (1.17-1.94)	1.15 (0.69-1.92)	1.00 (0.77-1.28)	1.54 (1.19-2.00)	1.09 (0.65-1.85)
South (Ponce)	1.15 (0.91-1.45)	1.12 (0.85-1.46)	1.31 (0.80-2.16)	1.26 (0.99-1.61)	1.18 (0.89-1.55)	1.26 (0.76-2.10)
Southeast (Caguas)	0.99 (0.78-1.26)	1.24 (0.95-1.61)	1.21 (0.73-2.02)	1.05 (0.82-1.35)	1.26 (0.96-1.65)	1.15 (0.69-1.93)

^*Multinomial logistic regression was performed using the low disease burden group—defined as individuals with 0–1 chronic condition—as the reference group.

^**Adjusted for all other variables in the table.

Table 3.

Associations Between Multimorbidity Patterns and Cancer Stage at Diagnosis and Tumor Location

	Crude Prevalence Ratio (95% CI)	Adjusted Prevalence Ratio (95% CI)*
Tumor location (Colon vs Rectum)
Multimorbidity Patterns
Low burden disease**	1.00	1.00
Hypertensive Vision – Joint	0.95 (0.93-0.97)	0.97 (0.95-1.00)
Diabetic Renal	0.93 (0.91-0.96)	0.95 (0.93-0.98)
Cardiopulmonary Severe	0.97 (0.92-1.02)	0.99 (0.94-1.05)
Clinical stage (Early vs Advanced)
Multimorbidity Patterns
Low burden disease**	1.00	1.00
Hypertensive Vision – Joint	0.88 (0.81-0.95)	0.94 (0.86-1.02)
Diabetic Renal	0.87 (0.79-0.95)	0.93 (0.85-1.02)
Cardiopulmonary Severe	0.83 (0.68-1.01)	0.90 (0.74-1.09)
Stage at diagnosis (Localized vs Distant/regional)
Multimorbidity Patterns
Low burden disease**	1.00	1.00
Hypertensive Vision – Joint	0.91 (0.85-0.97)	0.95 (0.89-1.01)
Diabetic Renal	0.88 (0.82-0.94)	0.93 (0.86-1.00)
Cardiopulmonary Severe	0.90 (0.78-1.03)	0.95 (0.82-1.09)

^*Adjusted for age, age², sex and insurance.

^**Individuals with 0–1 chronic condition.

Discussion

In this study, we characterized multimorbidity among adults diagnosed with CRC in Puerto Rico using claims-based chronic condition algorithms and latent class analysis. Over one-quarter of patients (28.6%) had two or more chronic conditions in addition to CRC, with hypertension, diabetes, and hyperlipidemia being the most prevalent. Latent class analysis identified three clinically interpretable multimorbidity profiles, Hypertensive Vision–Joint, Diabetic Renal, and Cardiopulmonary Severe, highlighting meaningful heterogeneity beyond comorbidity counts. Older age at diagnosis and insurance status were consistently associated with a higher likelihood of membership across these profiles, while sex and region were associated with specific patterns. After adjustment, latent class membership showed limited association with tumor location and stage at diagnosis, but no association with clinical stage. We observed small differences in tumor characteristics by multimorbidity profile. Specifically, the Hypertensive Vision–Joint and Diabetic Renal profiles showed slightly lower relative rates for tumor location compared with the reference class, and the Diabetic Renal profile showed a modestly lower relative rate for stage at diagnosis. Although these differences were statistically significant, effect sizes were small and should be interpreted cautiously.

Our findings indicate that multimorbidity is common among CRC patients in Puerto Rico and that cardiometabolic conditions account for a substantial share of the chronic disease burden in this population. This pattern is consistent with prior studies reporting that 40.7%–59% of cancer patients have one or more chronic conditions and identifying hypertension and diabetes as among the most prevalent comorbidities at diagnosis.^16,21,22 The consistency of these patterns across settings likely reflects shared underlying drivers, including aging and the high population prevalence of cardiometabolic risk factors, as well as increased healthcare contact among individuals with chronic disease.^10,21,34 Differences in multimorbidity prevalence across studies may also be explained by variation in study design and case definitions, including the number of chronic conditions assessed and whether multimorbidity was captured through registry data, self-report, or claims-based algorithms.^10,21,34

A key contribution of this study is the identification of three distinct multimorbidity profiles using latent class analysis, which may better capture clinically relevant disease clustering than traditional count-based measures. These profiles are consistent with disease clustering patterns reported in previous research conducted in the mainland United States and the United Kingdom.^12,13,22 Importantly, a systematic review of international studies concluded that the prognostic impact of multimorbidity depends more on the type and clustering of conditions than on the total number of conditions, supporting the clinical relevance of empirically derived profiles such as those identified here.¹⁴ The Diabetic Renal group resembles high-risk profiles reported in studies from Spain and Switzerland, where diabetes and chronic kidney disease commonly co-occur among CRC patients and have been associated with delayed care and poorer survival.^18,19 Similarly, our Cardiopulmonary Severe class aligns with cardiovascular and pulmonary disease clusters that have been linked to adverse outcomes in prior cohort studies.¹⁵ In addition, a study from China identified multiple latent multimorbidity classes among CRC patients, including clusters dominated by metabolic and cardiovascular conditions, reinforcing that these groupings may reflect common multimorbidity patterns across populations.²¹ A large integrated health system study in the United States also identified multimorbidity profiles associated with differential treatment receipt and survival, further supporting the utility of latent class models for CRC risk stratification.¹⁷

In our study, multimorbidity profile membership was patterned by sociodemographic characteristics, with older age, public insurance coverage, sex, and region showing meaningful associations. Older age at diagnosis was associated with a higher likelihood of membership in the Hypertensive Vision–Joint and Cardiopulmonary Severe classes, with a generally increasing pattern across age groups, consistent with evidence from population-based studies in Spain and Australia showing increasing comorbidity burden with age among CRC patients.^16,18 Insurance status also played a key role: dual Medicare–Medicaid coverage was disproportionately associated with membership in the Cardiopulmonary Severe profile, consistent with greater socioeconomic vulnerability and disease burden. Medicare-only coverage was also associated with this profile, likely reflecting older age and disability-related eligibility rather than socioeconomic disadvantage per se. These findings align with prior work indicating that older age, social deprivation, and barriers to healthcare access are predictors of chronic disease burden among cancer patients.¹³ A systematic review of 40 studies from Europe, North America, and Asia similarly highlighted structural vulnerabilities, including older age, lower income, and lack of private insurance, as key correlates of multimorbidity.¹⁴ We also observed sex differences, with women more likely to belong to the Hypertensive Vision–Joint class, consistent with evidence that the composition of multimorbidity differs by sex even when overall burden is similar.²⁰ Finally, regional differences in membership in the Diabetic Renal class may reflect variation in healthcare access and infrastructure across Puerto Rico’s health regions, including differences between urban and more rural municipalities, availability of specialty care, and patterns of healthcare utilization. Although Puerto Rico is geographically compact, regional disparities in socioeconomic conditions and healthcare resources may contribute to differences in chronic disease clustering.³⁵ Further research is needed to clarify these mechanisms.

These findings have implications for clinical practice and public health in Puerto Rico, particularly given the high prevalence of cardiometabolic conditions and the presence of distinct multimorbidity profiles among CRC patients. The identification of profiles such as Diabetic Renal and Cardiopulmonary Severe suggests that CRC care may benefit from integrated management strategies that address chronic disease complexity alongside cancer treatment planning, particularly for older adults and those with Medicare or dual Medicare–Medicaid coverage. Profile-based approaches may also support targeted survivorship planning and improved coordination between oncology and primary care. Future research should evaluate whether these latent classes are associated with differences in treatment receipt, treatment-related toxicity, healthcare utilization, and survival in Puerto Rico, and should explore how social determinants of health contribute to observed insurance and regional disparities in multimorbidity patterns. Future studies should also evaluate multimorbidity patterns among individuals with early-onset colorectal cancer, as disease burden, healthcare utilization, and treatment considerations may differ substantially in younger populations. Although individuals with multimorbidity typically have more frequent interactions with the healthcare system and may therefore have greater opportunities for earlier detection through contact with primary care providers, we did not observe a clear association with earlier stage at diagnosis. This finding raises important questions about preventive care delivery among patients with complex health needs. Competing clinical demands, fragmented care, or prioritization of chronic disease management over cancer screening may reduce opportunities for timely CRC detection. In addition, even when cancer is detected earlier, multimorbidity may complicate treatment decisions, delay treatment initiation, or limit treatment intensity, which may ultimately influence survival outcomes.

This study has several strengths, including the use of the PRCCR-HILD, which provides a comprehensive population-based sample of CRC patients in Puerto Rico, and the linkage of cancer registry data with health insurance claims, which improves ascertainment of chronic conditions. The use of latent class analysis is an additional strength, as it allows identification of empirically derived comorbidity profiles that may be more informative than traditional indices. However, limitations should be considered. Claims-based algorithms and condition-specific ascertainment windows may misclassify or under ascertain conditions that are less frequently coded or associated with lower healthcare utilization, potentially biasing prevalence estimates downward; this limitation may also differ by insurance type due to variation in healthcare utilization and coding practices. In addition, excluding individuals under 50 years of age reduces generalizability to early-onset CRC. Finally, although imputation methods were used for missing data, residual bias may remain.

Conclusion

In summary, multimorbidity was common among CRC patients in Puerto Rico, and latent class analysis identified three distinct profiles with potential clinical relevance. Older age and public insurance coverage were strongly associated with membership in these profiles, while sex and region were associated with specific patterns. These findings support the need for integrated cancer and chronic disease management strategies and provide a foundation for future research evaluating how multimorbidity profiles influence CRC treatment and outcomes in Puerto Rico.

Data Availability.

The data supporting the findings of this study were obtained from the PRCCR. Due to a confidentiality agreement between PRCCR and the authors, the clinical data used in this study are not publicly available. However, investigators may request access to the data through PRCCR by following the confidentiality procedures and submitting a request.