Comparing major comorbidity indices as predictors of all-cause mortality in the Veterans Affairs health care system.

Comparing major comorbidity indices as predictors of all-cause mortality in the Veterans Affairs healthcare system - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice J Clin Epidemiol . Author manuscript; available in PMC: 2026 Jun 1. Published in final edited form as: J Clin Epidemiol. 2025 Apr 2;182:111778. doi: 10.1016/j.jclinepi.2025.111778 Comparing major comorbidity indices as predictors of all-cause mortality in the Veterans Affairs healthcare system Hind A Beydoun a, b , Dorota Szymkowiak a , May A Beydoun c , Neil Nixdorff d , Robert Brunner e , Jack Tsai a, b, f Author information Article notes Copyright and License information a VA National Center on Homelessness Among Veterans, U.S. Department of Veterans Affairs, Washington, District of Columbia, United States b Department of Management, Policy, and Community Health, School of Public Health, University of Texas Health Science Center at Houston, Houston, Texas, United States c Laboratory of Epidemiology and Population Sciences, National Institute on Aging Intramural Research Program, Baltimore, Maryland, United States d Department of Internal Medicine, University of Michigan, Ann Arbor, Michigan, United States e Department of Family and Community Medicine (Emeritus), School of Medicine, University of Nevada (Reno), Nevada, United States f Department of Psychiatry, Yale School of Medicine, New Haven, Connecticut, United States ✉ Corresponding author : Hind A. Beydoun, PhD, MPH, National Center on Homelessness among Veterans, Veterans Health Administration; 810 Vermont Avenue, NW Washington, DC 20420; Hind.Baydoun@va.gov . Issue date 2025 Jun. PMC Copyright notice PMCID: PMC12146076  NIHMSID: NIHMS2071734  PMID: 40185292 The publisher's version of this article is available at J Clin Epidemiol Abstract Objective: The Charlson Comorbidity Index (CCI), the Elixhauser Comorbidity Index (ECI), and the Functional Comorbidity Index (FCI) are validated clinical measures of comorbidity, but direct comparisons between these measures have rarely been studied especially in high-risk patient populations, such as homeless individuals. The U.S. Department of Veterans Affairs (VA) offers large patient samples to compare these comorbidity measures as predictors of mortality using administrative and clinical records. We examined CCI, ECI, and FCI scores among veterans seeking VA healthcare services, including those experiencing homelessness, and compare their predictive value in relation to all-cause mortality risk. Study Design and Setting: Several VA databases from 2017–2021 were retrospectively linked and 4,701,711 U.S. veterans [308,553 with homelessness and 4,393,158 without homelessness] were evaluated over a median follow-up of 4.1 years, yielding 917,921 recorded deaths. Regression models were constructed, and Harrell’s Concordance Statistic (HCS) was calculated that assessed the ability of z-transformed comorbidity scores to discriminate ‘high-risk’ vs. ‘low-risk’ groups of patients for mortality risk, after adjustment for demographic and clinical characteristics. Results: In adjusted models, ECI (HCS: 0.76–0.77) and CCI (HCS: 0.75–0.76) were better able to discriminate ‘high-risk’ vs. ‘low-risk’ groups than FCI (HCS: 0.72–0.75) among homeless and non-homeless veterans. Compared to ECI and CCI, FCI was more strongly associated with homelessness. Conclusion: CCI and ECI may be more predictive of all-cause mortality risk than FCI, although FCI may be a useful measure of functioning in homeless populations. Keywords: Comorbidity, Epidemiology, Electronic Health Record, Homelessness, Mortality, Veteran Plain language summary: The Charlson Comorbidity Index (CCI), Elixhauser Comorbidity Index (ECI), and Functional Comorbidity Index (FCI) are clinical measures of comorbidity, but their direct comparisons have been limited, especially in high-risk patient populations like homeless individuals. This study examined CCI, ECI, and FCI scores among veterans seeking healthcare services at the U.S. Department of Veterans Affairs and found that ECI and CCI were better at discriminating high-risk vs. low-risk groups and predicting mortality among homeless and non-homeless veterans. 1. INTRODUCTION: Measures of comorbidity can be useful for risk prediction in clinical practice and research. Three comorbidity indices, namely, the Charlson Comorbidity Index (CCI) [ 1 – 6 ], the Elixhauser Comorbidity Index (ECI)[ 7 , 8 ], and the Functional Comorbidity Index (FCI)[ 9 – 11 ], have been developed, validated, and applied to a wide range of populations over the past 30 years. But they have rarely been compared in terms of their prediction of mortality risk. The CCI categorizes comorbidities based on International Classification of Diseases Clinical Modification (9 th and 10 th revisions) (ICD-9-CM/ICD-10-CM) diagnostic codes and total scores reflecting 1-year mortality risk and level of healthcare need[ 1 – 6 ]. The CCI originally covered 19 comorbidities[ 1 ] but was later modified to cover 17 comorbidities[ 2 ]. Originally developed with ICD-9-CM diagnostic codes, the ECI was later converted into ICD-10-CM diagnostic codes, with the number of ECI-specific comorbidities increasing from 29 to 38 over time[ 7 , 8 ]. The ECI score is used to predict in-hospital mortality and 30-day readmission risks[ 12 – 16 ]. Slightly lesser known than the CCI and the ECI, the FCI is comprised of 18 comorbidities and was initially developed as a self-administered questionnaire using a cross-sectional sample of 9,423 Canadian and 28,349 United States (U.S.) adults seeking spine treatment, suggesting a stronger relationship for FCI with physical function than with mortality risk indices such as the CCI[ 9 – 11 ]. The FCI was further developed to include ICD-9-CM/ICD-10-CM diagnostic codes and allow broader use of FCI in the context of administrative databases[ 11 ]. We found no prior published reports of a direct comparison of CCI, ECI, and FCI with routine electronic health records (EHRs) as the data source. The U.S. Department of Veterans Affairs (VA) operates the largest integrated healthcare system in the nation providing an opportunity for comparing these comorbidity indices in a large patient population. The VA uses a national EHRs system in which comorbidity scores can be calculated from ICD-9-CM/ICD-10-CM diagnostic codes. The VA’s large data system also allows comparison of these comorbidity scores among special populations including veterans with homelessness experience. Homelessness has been previously associated with multi-morbidities[ 17 , 18 ], as well as increased healthcare needs, morbidity, and mortality risks[ 16 , 19 – 26 ]. To date, CCI[ 27 – 31 ] and ECI[ 16 , 30 , 32 ] have rarely been compared as effective comorbidity indices in homeless populations and none have used FCI. In a recent study, Tsai and Kim analyzed 2019 Nationwide Inpatient Sample data and calculated ECI among homeless and non-homeless hospitalized patients[ 16 ]. Their results suggested that schizophrenia, depressive disorders, and alcohol-related disorders were leading diagnoses in the homeless group, metastatic cancer and liver disease were leading diagnoses in the combined groups, and ECI was predictive of in-hospital mortality risk supporting its use as a case-mix control method[ 16 ]. Although ECI has been previously shown to outperform CCI in a variety of clinical populations[ 13 , 30 , 33 , 34 ], to date, no studies have compared the performances of CCI, ECI, and FCI as predictors of all-cause mortality risk, taking homelessness status into consideration. The main objectives of this study were to (1) examine CCI, ECI, and FCI scores among a large sample of older patients using data from the VA healthcare system, including those who have experienced homelessness; and (2) compare the predictive abilities of these three comorbidity indices in relation to all-cause mortality risk among patients in the VA healthcare system, overall, and by homelessness status. We hypothesized that the ECI would be a better predictor of all-cause mortality than the CCI or FCI. 2. METHODS: 2.1. Databases: Three VA databases – Corporate Data Warehouse (CDW), Homeless Operations Management System (HOMES), and Department of Defense [DoD]/VA Joint Mortality Data Repository (MDR) – were linked using Microsoft’s Structured Query Language Management Studio version 18 (Redmond, WA) and Statistical Analysis System Enterprise Guide version 8.3 (SAS Institute, Cary, NC) on the VA’s Informatics and Computing Infrastructure servers behind the VA firewall. The CDW is VA’s national repository of EHRs comprising data on all inpatient and outpatient encounters from the VA administrative and clinical systems.[ 35 , 36 ] Data extracted from the CDW databases consist of patient demographics, diagnoses, and procedures, including dates and settings of healthcare use[ 35 , 37 – 39 ]. Populated by homeless service staff, the HOMES database contains information on veteran participation in almost all VA homeless programs. The DoD/VA MDR contains mortality data from the National Death Index (NDI), including date of death, cause of death, state of death, date of birth, and sex of the decedent. This project conducted under the auspices of the VA Homeless Programs Office for quality improvement did not constitute human subjects research and was exempt from review by an institutional review board. See attached STROBE Checklist. 2.2. Study population: Using the linked CDW and HOMES databases, we defined a cohort of older veterans aged 50 years or older at their index visit, i.e., initial visit to the VA healthcare system between 1 January 2017 and 31 December 2018. This cohort of veterans was followed between their index visit and 31 December 2021. Mortality data from the MDR database were linked with this cohort of veterans to define the final analytic sample. 2.3. Measures: 2.3.1. Homelessness: We used a VA operational definition of homelessness[ 40 ] to identify veterans who experienced homelessness anytime between 1 January 2017 and 31 December 2018. More specifically, veterans were defined as homeless if they had an ICD-10-CM diagnostic code descriptive of homelessness (Z59.0); service stop codes or HOMES records indicating clinical encounters with homeless service staff in the Grant and Per Diem program, the Health Care for Homeless Veterans program, the Domiciliary Care for Homeless Veterans program, or the Compensated Work Therapy/Transitional Residence program; or had a positive screen on an annual VA homeless screening clinical reminder. 2.3.2. Comorbidity indices: CCI[ 1 – 6 ], ECI[ 7 , 8 ], and FCI[ 9 – 11 ] scores were calculated using inpatient and outpatient ICD-10-CM diagnoses recorded within 12 months after each veteran’s index visit. As previously reported in the literature, for the 17-item CCI[ 6 ] and 18-item FCI[ 11 ], ICD-10-CM diagnoses were applied for the purpose of comorbidity scoring. Of note, we used 38 comorbidities and the 2024 version of the ‘Elixhauser comorbidity software refined for ICD-10-CM index program’ developed by the Agency for Healthcare Research and Quality Healthcare Cost and Utilization Project to calculate ECI scores whereby weights that reflect risk of in-hospital mortality ( mw ) rather than 30-day all-cause readmission ( rw ) were applied[ 8 , 41 ]. Continuous CCI, ECI, and FCI scores were z-scored or categorized using restricted natural splines with three knots as a flexible tool to evaluate relationships with all-cause mortality risk. 2.3.3. All-cause mortality risk: The risk of dying between the index visit and 31 December 2021 was defined using two data elements from the MDR database, namely, a dichotomous (Yes/No) variable indicating death status and date of death among those who died. Follow-up time was calculated as time duration from index visit until death from any cause, loss to follow-up, or end of follow-up, as applicable. Specifically, veterans surviving until 31 December 2021 were assigned their maximum follow-up time from their index visit. 2.3.4. Demographic and clinical characteristics: Hypothesized relationships were examined before and after adjustment for sex, age, race, marital status, combat service, and percent service-connected disability rating at each veteran’s index visit between 1 January 2017 and 31 December 2018, in addition to substance use ( i.e., tobacco, alcohol, drugs), and mental health ( i.e., mood, psychotic, anxiety, stress, personality, other) disorders assessed within 12 months after each veteran’s index visit. 2.4. Statistical analysis: Descriptive, bivariate, and multivariable analyses were performed using the SAS Entreprise Guide version 8.3. First, the demographic and clinical characteristics were examined by homelessness status. Second, comorbidity scores were described by demographic and clinical characteristics, before and after stratifying by homelessness status. Third, multivariable Cox regression models were constructed for demographic and clinical characteristics as predictors of all-cause mortality risk, before and after stratifying by homelessness status. Fourth, we constructed logistic regression models to examine z-scored comorbidity indices as predictors of homelessness status, before and after adjustment for demographic and clinical characteristics. Finally, we constructed Cox regression models to examine each comorbidity score – z-scored or categorized using restricted natural splines – as a predictor of all-cause mortality risk, by homelessness status, before and after adjustment for demographic and clinical characteristics. To address immortal time bias, we also constructed multivariable Cox regression models to examine z-scored comorbidity indices as predictors of all-cause mortality risk among a subset of homeless and non-homeless veterans who survived beyond 12 months of follow-up. Logistic and Cox regression models were applied to estimate unadjusted and adjusted odds ratios (OR) and hazard ratios (HR) with their 95% confidence intervals (CI), respectively. Previous studies have suggested a wide range of performance measures to be applied in the context of Cox regression modeling[ 12 , 42 , 43 ]. Using the CONCORDANCE option within the PHREG procedure, we estimated the Harrell’s Concordance Statistic (HCS) to assess the abilities of comorbidity scores to discriminate ‘high-risk’ vs. ‘low-risk’ groups of veterans on all-cause mortality risk. The HCS is interpreted in a similar fashion to area under the receiver operating curve with HCS values close to 0.5 indicating predictions were not better than chance in determining which patient will be at greater risk of dying, and HCS values close to 1 indicating predictions were better at determining which patients will be at greater risk of dying. Two-sided statistical tests were performed assuming α=0.05. 3. RESULTS: Of 6,347,780 veterans who sought VA healthcare services between 1 January 2017 and 31 December 2018, 6,299,622 (99.2%) were successfully linked using CDW, HOMES, and MDR databases. Of those with linked data, 4,709,824 (74.8%) were aged 50–100 years at their index visits. The analytic sample consisted of 4,701,711 (99.8%) veterans – 308,553 (6.6%) homeless and 4,393,158 (93.4%) non-homeless – after excluding those with invalid data on demographic, clinical, and mortality characteristics ( Appendix A , Figure A.1. ). The median [interquartile range] follow-up times between index visit and 31 December 2021 for the full sample, the homeless, and non-homeless subgroups, were 1,704 [1,487–1,788], 1,730 [1,352–1,801], and 1,703 [1,492–1,787] days, respectively, yielding 917,921 deaths – 84,982 among homeless and 823,939 among non-homeless veterans. As shown in Table 1 , homelessness-experienced veterans were younger, more frequently female, non-white, and unmarried, and more likely to be diagnosed with psychiatric disorders, but less likely to have received a VA service-connected disability rating or to report combat service than veterans with no homelessness experience. Table 1. Demographic and clinical characteristics by homelessness status (n=4,701,711) Homelessness Yes (N=308,553) No (N=4,393,158) Overall (N=4,701,711) N (%) or Mean (SD) N (%) or Mean (SD) N (%) or Mean (SD) Sex: P<0.001 Male 289,433 (93.8) 4,157,391 (94.6) 4,446,824 (94.6) Female 19,120 (6.2) 235,767 (5.4) 254,887 (5.4) Age (years): P<0.001 Mean ± SD 65.0 ± 9.5 69.5 ± 10.3 69.2 ± 10.3 P<0.001 50–59 107,319 (34.8) 848,457 (19.3) 955,776 (20.3) 60–69 123,198 (39.9) 1,576,794 (35.9) 1,699,992 (36.2) 70–79 50,470 (16.4) 1,219,596 (27.8) 1,270,066 (27.0) ≥ 80 27,566 (8.9) 748,311 (17.0) 775,877 (16.5) Race: P<0.001 Non-Hispanic White 165,027 (53.5) 3,109,913 (70.8) 3,274,940 (69.7) Non-Hispanic Black 96,649 (31.3) 628,866 (14.3) 725,515 (15.4) Hispanic 17,893 (5.8) 173,232 (3.9) 191,195 (4.1) Mixed race / Other 7,265 (2.4) 107,150 (2.4) 114,415 (2.4) Missing 21,719 (7.0) 373,997 (8.5) 395,716 (8.4) Marital status: P<0.001 Married 87,439 (28.3) 2,646,030 (60.2) 2,733,469 (58.1) Single / Never married 57,293 (18.6) 345,845 (7.9) 403,138 (8.6) Separated / Divorced 138,944 (45.0) 1,038,153 (23.6) 117,097 (25.0) Widowed 23,202 (7.5) 325,992 (7.4) 349,194 (7.4) Missing 1,675 (0.5) 37,138 (0.9) 38,813 (0.8) % service-connected disability rating: P<0.001 None / 0% 154,200 (49.9) 1,940,806 (44.2) 2,095,006 (44.6) 10%-40% 48,729 (15.8) 779,707 (17.8) 828,436 (17.6) 50%-100% 105,624 (34.2) 1,672,645 (38.1) 1,778,269 (37.8) Combat service: P<0.001 Yes 16,081 (5.2) 306,492 (6.9) 322,573 (6.9) No 292,472 (94.8) 4,086,666 (93.0) 4,379,138 (93.1) Substance use disorder: P<0.001 Yes 101,568 (32.9) 309,573 (7.0) 411,141 (8.7) No 206,985 (67.1) 4,083,585 (92.9) 4,290,570 (91.3) Mental health disorder: P<0.001 Yes 128,903 (41.8) 744,396 (16.9) 873,299 (18.6) No 179,650 (58.2) 3,648,762 (83.1) 3,828,412 (81.4) Open in a new tab Abbreviations: SD = Standard deviation. Table 2 presents comorbidity scores by demographic and clinical characteristics, overall, and after stratifying by homelessness status. Overall, the mean [±standard deviation (SD)] CCI, ECI, and FCI scores were 0.55 [±1.17] (range: 0–17), −3.53 [±8.43] (range: −51–120), and 1.43 [±1.62] (range: 0–14), respectively. On average, comorbidity scores were higher among veterans with vs. without homelessness experiences. General linear models for comorbidity scores indicated significant differences according to background characteristics, before and after stratifying by homelessness status. It is worth noting that, on average, CCI and ECI scores were higher among male veterans, while FCI scores were, on average, higher among female veterans. In addition, all comorbidity index scores increased with age, regardless of homelessness status. Table 2. Comorbidity index scores by demographic and clinical characteristics and homelessness status (n=4,701,711) * CCI Overall (n=4,701,711) Mean ± SD ECI FCI CCI Homelessness [Yes] (n=308,553) Mean ± SD ECI FCI Homelessness [No] (n=4,393,158) Mean ± SD CCI ECI FCI Overall: 0.55±1.17 −3.53±8.43 1.43±1.62 1.18±1.77 −2.89±14.49 2.41±2.01 0.51±1.10 −3.57±7.83 1.37±1.57 Sex: Male 0.56±1.19 −3.35±8.43 1.43±1.62 1.21±1.79 −2.51±14.57 2.40±2.01 0.52±1.12 −3.40±7.82 1.36±1.56 Female 0.34±0.86 −6.70±7.93 1.52±1.67 0.68±1.28 −8.58±11.78 2.53±1.97 0.31±0.81 −6.55±7.52 1.43±1.61 Age (years): 50–59 0.31±0.93 −6.06±7.71 1.28±1.53 0.62±1.33 −8.20±11.68 2.03±1.82 0.27±0.86 −5.79±7.00 1.18±1.46 60–69 0.56±1.17 −3.98±8.62 1.58±1.68 1.19±1.74 −2.93±14.14 2.47±2.04 0.52±1.10 −4.06±8.03 1.51±1.63 70–79 0.63±1.23 −2.56±8.14 1.45±1.63 1.76±1.98 2.60±14.99 2.82±2.12 0.58±1.16 −2.78±7.65 1.39±1.58 ≥ 80 0.69±1.28 −0.99±8.39 1.29±1.55 2.23±2.08 7.90±15.19 2.89±2.06 0.63±1.22 −1.32±7.84 1.24±1.49 Race: Non-Hispanic White 0.54±1.13 −3.39±8.18 1.43±1.63 1.25±1.75 −1.96±14.74 2.61±2.06 0.49±1.08 −3.47±7.67 1.37±1.58 Non-Hispanic Black 0.69±1.39 −3.79±9.81 1.50±1.59 1.10±1.81 −4.57±14.07 2.11±1.87 0.62±1.30 −3.67±8.97 1.41±1.53 Hispanic 0.56±1.21 −4.04±9.01 1.62±1.66 1.33±1.88 −1.44±15.27 2.62±2.03 4.48±1.09 −4.31±8.04 1.51±1.58 Mixed race / Other 0.46±1.04 −3.99±7.77 1.37±1.56 0.85±1.49 −4.15±12.48 2.07±1.92 0.43±1.00 −3.98±7.34 1.33±1.52 Missing 0.45±1.05 −3.74±7.64 1.24±1.53 1.01±1.62 −3.21±13.54 2.19±1.98 0.42±1.00 −3.77±7.14 1.19±1.49 Marital status: Married 0.49±1.08 −3.49±7.72 1.41±1.61 1.51±1.89 0.28±15.44 2.88±2.09 0.46±1.03 −3.61±7.29 1.36±1.57 Single / Never married 0.60±1.35 −4.33±9.33 1.32±1.53 0.89±1.66 −5.57±12.86 1.89±1.80 0.55±1.29 −4.13±8.59 1.22±1.46 Separated / Divorced 0.62±1.25 −3.83±9.37 1.53±1.65 1.01±1.64 −4.49±13.76 2.28±1.94 0.57±1.18 −3.74±8.62 1.43±1.58 Widowed 0.74±1.34 −1.89±9.25 1.53±1.65 1.65±1.96 1.31±15.71 2.74±2.08 0.67±1.25 −2.11±8.57 1.44±1.62 Missing 0.31±0.94 −3.72±6.75 0.87±1.36 1.03±1.85 −2.12±14.13 1.94±1.99 0.28±.087 −3.79±6.20 0.82±1.30 % service-connected disability rating: None / 0% 0.56±1.20 −2.76±8.35 1.16±1.46 1.18±1.78 −1.76±14.31 2.06±1.88 0.51±1.13 −2.85±7.67 1.09±1.39 10%-40% 0.43±1.04 −3.63±7.36 1.19±1.46 1.02±1.65 −3.69±13.68 2.25±1.93 0.39±0.97 −3.63±6.77 1.13±1.40 50%-100% 0.59±1.19 −4.37±8.91 1.87±1.77 1.25±1.79 −4.16±14.97 3.00±2.08 0.55±1.13 −4.39±8.38 1.79±1.73 Combat service: Yes 0.64±1.24 −3.15±8.89 1.73±1.75 1.54±1.94 0.05±15.55 2.96±2.13 0.59±1.18 −3.32±8.37 1.67±1.70 No 0.54±1.17 −3.56±8.39 1.41±1.61 1.16±1.76 −3.05±14.41 2.38±1.99 0.50±1.09 −3.59±7.79 1.34±1.55 Substance use disorder: Yes 1.08±1.61 −5.01±13.84 2.58±1.82 1.28±1.82 −6.46±16.46 2.93±1.93 1.01±1.52 −4.54±12.83 2.47±1.77 No 0.50±1.11 −3.38±7.70 1.32±1.55 1.13±1.74 −1.14±13.06 2.16±1.99 0.47±1.05 −3.49±7.31 1.28±1.52 Mental health disorder: Yes 0.86±1.43 −8.10±11.50 3.16±1.71 1.27±1.80 −8.17±15.24 3.51±1.91 0.79±1.34 −8.09±10.72 3.11±1.66 No 0.48±1.09 −2.48±11.50 1.04±1.31 1.12±1.74 0.89±12.63 1.63±1.68 0.44±1.04 −2.65±6.74 1.01±1.28 Open in a new tab Abbreviations: CCI = Charlson comorbidity index; ECI = Elixhauser comorbidity index; FCI = Functional comorbidity index; SD = Standard deviation. * Independent samples t-tests suggested highly significant differences between homeless and non-homeless veterans on CCI, ECI, and FCI scores, with P < 0.001. General linear models for CCI, ECI, and FCI scores in relation to each demographic or clinical characteristic resulted in highly significant differences by level of these characteristics, overall, as well as among homeless and non-homeless veterans, with P < 0.001. Multivariable Cox regression models for relationships of demographic and clinical characteristics with all-cause mortality risk by homelessness status are displayed in Table 3 . Overall, male sex, age, unmarried status, combat service, and psychiatric diagnoses were associated with greater risk of dying and >0% service-connected disability rating was associated with lower risk of dying within the follow-up period. Veterans who were not homeless had greater risk and those who were homeless had lower risk of dying if they reported their race as non-Hispanic black vs. non-Hispanic white. Similarly, risk of dying was higher among unmarried non-homeless veterans, but lower among unmarried homeless veterans. Table 3. Multivariable cox regression models for the relationship of demographic and clinical characteristics with all-cause mortality risk by homelessness status (n=4,701,711) All-cause mortality risk Homelessness [Yes] (n=308,553) Homelessness [No] (n=4,393,158) Overall (n=4,701,711) HR (95% CI) HR (95% CI) HR (95% CI) Sex: Male 1.65 (1.59, 1.72) 1.64 (1.61, 1.67) 1.65 (1.63, 1.68) Female Ref. Ref. Ref. Age (years): Continuous 1.07 (1.07, 1.07) 1.09 (1.09, 1.09) 1.09 (1.09, 1.09) Race: Non-Hispanic White Ref. Ref. Ref. Non-Hispanic Black 0.80 (0.79, 0.81) 1.02 (1.01, 1.03) 1.03 (1.02, 1.03) Hispanic 0.79 (0.77, 0.82) 0.78 (0.77, 0.79) 0.82 (0.81, 0.83) Mixed race / Other 0.86 (0.82, 0.89) 0.92 (0.90, 0.93) 0.92 (0.91, 0.94) Missing 0.79 (0.77, 0.82) 0.82 (0.81, 0.83) 0.82 (0.81, 0.83) Marital status: Married Ref. Ref. Ref. Single / Never married 0.86 (0.84, 0.88) 1.33 (1.32, 1.34) 1.34 (1.33, 1.35) Separated / Divorced 0.90 (0.88, 0.92) 1.32 (1.31, 1.33) 1.33 (1.32, 1.34) Widowed 0.92 (0.89, 0.94) 1.13 (1.12, 1.14) 1.14 (1.14, 1.15) Missing 1.24 (1.13, 1.35) 1.38 (1.34, 1.42) 1.39 (1.36, 1.43) % service-connected disability rating: None / 0% Ref. Ref. Ref. 10%-40% 0.82 (0.80, 0.84) 0.78 (0.78, 0.79) 0.78 (0.78, 0.79) 50%-100% 0.73 (0.72, 0.75) 0.78 (0.78, 0.79) 0.77 (0.76, 0.77) Combat service: Yes 1.16 (1.13, 1.19) 1.13 (1.12, 1.14) 1.13 (1.12, 1.13) No Ref. Ref. Ref. Substance use disorder: Yes 1.42 (1.39, 1.44) 1.74 (1.72, 1.75) 1.82 (1.81, 1.83) No Ref. Ref. Ref. Mental health disorder: Yes 1.09 (1.07, 1.11) 1.28 (1.27, 1.29) 1.31 (1.31, 1.32) No Ref. Ref. Ref. Open in a new tab Abbreviations: CI = Confidence interval; HR = Hazard ratio. Table 4 presents logistic regression models for z-scored comorbidity indices in relation to homelessness experience, suggesting that the relationship of homelessness with z-scored FCI was strongest (adjusted OR=1.52, 95% CI: 1.51, 1.53), followed by z-scored CCI (adjusted OR=1.36, 95% CI: 1.36, 1.36), and z-scored ECI (adjusted OR=1.25, 95% CI: 1.25, 1.25), after adjusting for demographic and clinical characteristics. Table 4. Logistic regression models for the relationships of z-scored comorbidity indices with homelessness experience, before and after controlling for demographic and clinical characteristics (n=4,701,711) * Odds Ratio (95% Confidence Interval) Unadjusted Adjusted † CCI: 1.44 (1.44, 1.45) 1.36 (1.36, 1.36) ECI: 1.08 (1.08, 1.08) 1.25 (1.25, 1.25) FCI: 1.66 (1.66, 1.66) 1.52 (1.51, 1.53) Open in a new tab Abbreviations: CCI = Charlson comorbidity index; ECI = Elixhauser comorbidity index; FCI = Functional comorbidity index. * Odds ratios with their 95% confidence intervals were estimated for the relationship of 1 standard deviation change in each of CCI, ECI, and FCI with homelessness experience as the outcome variable within logistic regression models. † Adjusted for sex, age, race, marital status, combat service, percent service-connected disability rating, substance use ( i.e ., tobacco, alcohol, drugs), and mental health ( i.e ., mood, psychotic, anxiety, stress, personality, other) characteristics. Cox regression models for z-scored comorbidity indices as predictors of all-cause mortality risk, are displayed by homelessness experience, before and after adjustment for demographic and clinical characteristics ( Table 5 ). The z-scored ECI (unadjusted HR=1.47, HCS=0.70; adjusted HR=1.37, HCS=0.76) and z-scored CCI (unadjusted HR=1.41, HCS=0.69; adjusted HR=1.33, HCS=0.75) were more strongly related to all-cause mortality risk and better able to discriminate ‘high-risk’ vs. ‘low-risk’ groups than z-scored FCI (unadjusted HR=1.25, HCS=0.58; adjusted HR=1.19, HCS=0.72) among homeless-experienced veterans. Similarly, z-scored ECI (unadjusted HR=1.56, HCS=0.63; adjusted HR=1.43, HCS=0.77) and z-scored CCI (unadjusted HR=1.43, HCS=0.60; adjusted HR=1.36, HCS=0.76) were more strongly related to all-cause mortality risk and better able to discriminate ‘high-risk’ vs. ‘low-risk’ groups than z-scored FCI (unadjusted HR=1.16, HCS=0.53; adjusted HR=1.19, HCS=0.75) among veterans without homelessness experience. In adjusted models, statistically significant two-way interaction terms were observed for CCI-by-homelessness status (β=+0.05, SE=0.001, P<0.001), ECI-by-homelessness status (β=+0.01, SE=0.002, P<0.001), and FCI-by-homelessness status (β=+0.15, SE=0.002, P<0.001). Similar results were obtained after excluding veterans who died within 12 months of follow-up ( Appendix A , Table A.1. ). Table 5. Cox regression models for the relationship of z-scored comorbidity indices with all-cause mortality risk, overall, and by homelessness status, before and after controlling for demographic and clinical characteristics (n=4,701,711) * Hazard ratio (95% Confidence Interval) Harrell’s Concordance Statistic Unadjusted Adjusted † HOMELESSNESS – YES (n=308,553): CCI: 1.41 (1.41, 1.41) 0.69 1.33 (1.32, 1.33) 0.75 ECI: 1.47 (1.47, 1.48) 0.70 1.37 (1.37, 1.37) 0.76 FCI: 1.25 (1.25, 1.26) 0.58 1.19 (1.18, 1.19) 0.72 HOMELESSNESS – NO (n=4,393,158): CCI: 1.43 (1.43, 1.43) 0.60 1.36 (1.36, 1.36) 0.76 ECI: 1.56 (1.56, 1.56) 0.63 1.43 (1.43, 1.43) 0.77 FCI: 1.16 (1. 16, 1.16) 0.53 1.19 (1.19, 1.19) 0.75 Open in a new tab Abbreviations: CCI = Charlson comorbidity index; ECI = Elixhauser comorbidity index; FCI = Functional comorbidity index; SE = Standard error; * Hazard ratios with their 95% confidence intervals were estimated for the relationship of 1 standard deviation change in each of CCI, ECI, and FCI with all-cause mortality risk as the outcome variable within Cox regression models. In adjusted models, statistically significant two-way interaction terms were observed for CCI-by-homelessness status (β=+0.05, SE=0.001, P<0.001), ECI-by-homelessness status (β=+0.01, SE=0.002, P<0.001), and FCI-by-homelessness status (β=+0.15, SE=0.002, P<0.001); † Adjusted for sex, age, race, marital status, combat service, percent service-connected disability rating, substance use ( i.e ., tobacco, alcohol, drugs), and mental health ( i.e ., mood, psychotic, anxiety, stress, personality, other) characteristics. After identifying break points for spline effects ( Appendix A , Table A.2. ), a positive dose-response relationship was observed between two comorbidity indices (CCI and ECI), defined as categorical variables, and all-cause mortality risks, before and after stratifying by homelessness status ( Appendix A , Table A.3. ). Conversely, the relationship between FCI and all-cause mortality risk varied according to homelessness status, with a stronger relationship observed among homeless veterans than their counterparts. 4. DISCUSSION: The CCI[ 1 – 6 , 44 – 46 ], ECI[ 7 , 8 ], and FCI[ 9 – 11 ], are three comorbidity indices that are commonly used for the purpose of prognostication, in clinical practice and research, but their abilities to predict all-cause mortality risk have not been directly compared. The comprehensive VA EHRs data offered a unique opportunity to compare these comorbidity indices. In this retrospective cohort study, we analyzed CCI, ECI, and FCI scores among 4,701,711 U.S. veterans seeking VA healthcare services, including those experiencing homelessness, yielding a total of 917,921 deaths over ~4.1 years of follow-up. All three comorbidity indices were reasonably good predictors of all-cause mortality risk as evidenced by their HSC scores, which ranged between 0.53 and 0.69 for unadjusted models, and between 0.72 and 0.77 for models that adjusted for sex, age, race, marital status, combat service, percent service-connected disability rating, substance use and mental health disorders. Notably though, the ECI and CCI scores were more strongly related to all-cause mortality risk and better able to discriminate high-risk vs. low-risk groups than the FCI. This was true for both homeless and non-homeless veterans, although ECI and CCI were more strongly related to all-cause mortality risk among non-homeless than homeless patients. As such, ECI and CCI may be better indices for use when assessing risk of death among veterans. Conversely, FCI may be useful in other contexts including the assessment of physical functioning and rehabilitation healthcare needs of veterans. Results showed that homelessness was more strongly related to the FCI than the ECI and CCI, implying that homelessness may be more closely linked to functional deficits than solely with comorbid conditions. It is worth noting that compared to veterans with FCI = 0, those with FCI ≥ 2 had greater all-cause mortality risk if they had experienced homelessness than if they had not experienced homelessness. This unexpected finding highlights the vulnerability of homelessness-experienced veterans to poor prognosis after being diagnosed with chronic conditions known to adversely affect physical functioning and lead to disability. Recent studies of a diverse range of populations have examined abilities of comorbidity indices, especially the ECI and CCI, to predict health-related outcomes among older populations. For instance, a study aimed to develop and validate a parsimonious model for predicting 30-day hospital readmission from home healthcare [HHC] in Medicare fee-for-service beneficiaries.[ 47 ] Using multivariable logistic regression and gradient boosting machine learning on a 5% national sample of patients ≥ 65 years of age who received skilled HHC services within 5 days of hospital discharge in 2012, seven candidate variables remained in the final model, including ECI, with c-statistics > 0.60.[ 47 ] Another study involving 555 patients aimed to develop and validate a prognostic model for 6-month and 1-year mortality risks in home-dwelling patients with advanced dementia, whereby the model was validated using Cox regression and further refined to include CCI among other covariates.[ 48 ] Finally, a 9-year study aimed at assessing premature cognitive aging and its relationship with early mortality risk in homeless and precariously housed individuals found that for each 1-point increase in CCI score, mortality risk was 9.9% higher[ 27 ]. Our study has several notable strengths. Linked databases covering homelessness, comorbidity scores, mortality, and covariate data, enabled the unique contribution of this study. With > 4,000,000 veterans, our study had among the largest sample sizes in the literature on this topic. Nevertheless, study findings should be interpreted taking several limitations into account. First, routinely collected data were analyzed limiting the study’s ability to evaluate additional confounding variables not collected within VA databases. Second, missing data on background characteristics may have resulted in selection bias. Third, misclassification bias may have resulted from use of self-report, diagnostic and stop codes to define demographic, clinical and homelessness characteristics. Homelessness varies over time and may differ for each veteran before, during, and after the study’s observation period. Fourth, causal relationships between comorbidity indices and all-cause mortality risk cannot be established with an observational study design. Fifth, statistical significance may not necessarily reflect clinical significance in the context of large sample sizes, and especially when examining the proportional hazards assumption and interaction effects.[ 49 , 50 ] Finally, study findings can only be generalized to veterans who sought VA healthcare services and were followed-up between 2017 and 2021. In conclusion, our analyses using VA linked databases suggested that the CCI and ECI were better predictors of all-cause mortality risks than the FCI. The FCI was more strongly related to homelessness than either CCI or ECI. Further research is needed to elucidate mechanisms by which the CCI and ECI were more predictive of all-cause death, and in what contexts the FCI may be a useful assessment tool for various populations including homeless veterans. The usefulness of comorbidity indices as tools for determining resource allocation in the context of veteran homelessness requires further investigation. Supplementary Material 1 NIHMS2071734-supplement-1.docx (85.6KB, docx) Highlights: Performance of CCI, ECI, and FCI have rarely been compared. Study compared CCI, ECI, and FCI among >4 million veterans on predicting mortality. ECI and CCI better predicted mortality risk among all veterans than FCI. FCI was more strongly associated with homelessness than ECI or CCI. What is New? 1. Key findings: A study of over 4 million U.S. veterans found that two comorbidity indices (ECI and CCI) were more effective than the FCI in distinguishing high-risk vs. low-risk groups for mortality among veterans. 2. What this adds to what is known? This study identifies among the CCI, ECI, and FCI, which indices are better predictors of mortality in homeless and non-homeless veterans. 3. What is the implication, what should change now? CCI and ECI may be more useful in predicting mortality risk than FCI, although FCI may be useful in assessing functioning among homeless populations. Acknowledgments: We thank the Department of Veterans Affairs, the Homeless Programs Office and the Office of Mental Health and Suicide Prevention for their support of this study. The Joint Department of Veterans Affairs and Department of Defense Mortality Data Repository were provided by the Department of Veterans Affairs, Office of Mental Health and Suicide Prevention. These data were compiled from the National Death Index. https://www.mirecc.va.gov/suicideprevention/documents/VA_DoD-MDR_Flyer.pdf . Sources of funding: This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Footnotes Declaration of interests ☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Disclaimer : The views expressed in this article are those of the authors and do not reflect the official policy of the U.S. Department of Veterans Affairs, the U.S. National Institutes of Health, or the U.S. government. Credit Roles: Conceptualization: HB, JT, RB Data curation: HB, DS Formal analysis: HB Funding acquisition: N/A Investigation: HB, DS, JT Methodology: HB, DS, JT Project administration: HB, JT Resources: JT Software: HB, MB Supervision: JT Validation: MB, NN, RB Visualization: HB Writing – original draft: HB Writing – review and editing: MB, NN, RB, JT Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. LIST OF APPENDICES APPENDIX A: CDW_HUD_VASH_MANUSCRIPT_3_CCI_ECI_FCI_APPENDA_DRAFT10.DOC Note: The Appendix includes Figure A.1 , Table A.1 , Table A.2. , and Table A.3. Data Availability Statement: The data that support the findings of this study are available from the U.S. Department of Veterans Affairs. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the authors with the permission of the U.S. Department of Veterans Affairs. References: 1. Charlson ME, Pompei P, Ales KL, MacKenzie CR: A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987, 40(5):373–383. [ DOI ] [ PubMed ] [ Google Scholar ] 2. Deyo RA, Cherkin DC, Ciol MA: Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992, 45(6):613–619. [ DOI ] [ PubMed ] [ Google Scholar ] 3. Romano PS, Roos LL, Jollis JG: Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J Clin Epidemiol 1993, 46(10):1075–1079; discussion 1081–1090. [ DOI ] [ PubMed ] [ Google Scholar ] 4. Schneeweiss S, Wang PS, Avorn J, Glynn RJ: Improved comorbidity adjustment for predicting mortality in Medicare populations. Health Serv Res 2003, 38(4):1103–1120. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Halfon P, Eggli Y, van Melle G, Chevalier J, Wasserfallen JB, Burnand B: Measuring potentially avoidable hospital readmissions. J Clin Epidemiol 2002, 55(6):573–587. [ DOI ] [ PubMed ] [ Google Scholar ] 6. Sundararajan V, Henderson T, Perry C, Muggivan A, Quan H, Ghali WA: New ICD-10 version of the Charlson comorbidity index predicted in-hospital mortality. J Clin Epidemiol 2004, 57(12):1288–1294. [ DOI ] [ PubMed ] [ Google Scholar ] 7. Elixhauser A, Steiner C, Harris DR, Coffey RM: Comorbidity measures for use with administrative data. Med Care 1998, 36(1):8–27. [ DOI ] [ PubMed ] [ Google Scholar ] 8. Moore BJ, White S, Washington R, Coenen N, Elixhauser A: Identifying Increased Risk of Readmission and In-hospital Mortality Using Hospital Administrative Data: The AHRQ Elixhauser Comorbidity Index. Med Care 2017, 55(7):698–705. [ DOI ] [ PubMed ] [ Google Scholar ] 9. Agrafiotis M, Galanou A, Fletsios D, Chassiotou A, Chloros D, Steiropoulos P: Functional Comorbidity Index and health-related quality of life in patients with obstructive sleep apnea. Adv Respir Med 2022. doi: 10.5603/ARM.a2022.0001. [ DOI ] [ PubMed ] [ Google Scholar ] 10. Groll DL, To T, Bombardier C, Wright JG: The development of a comorbidity index with physical function as the outcome. J Clin Epidemiol 2005, 58(6):595–602. [ DOI ] [ PubMed ] [ Google Scholar ] 11. Sears JM, Rundell SD: Development and Testing of Compatible Diagnosis Code Lists for the Functional Comorbidity Index: International Classification of Diseases, Ninth Revision, Clinical Modification and International Classification of Diseases, 10th Revision, Clinical Modification. Med Care 2020, 58(12):1044–1050. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 12. Bajic B, Galic I, Mihailovic N, Ristic S, Radevic S, Cupic VI, Kocic S, Arnaut A: Performance of Charlson and Elixhauser Comorbidity Index to Predict in-Hospital Mortality in Patients with Stroke in Sumadija and Western Serbia. Iran J Public Health 2021, 50(5):970–977. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Bonnesen K, Mols RE, Logstrup B, Gustafsson F, Eiskjaer H, Schmidt M: The Ability of Comorbidity Indices to Predict Mortality After Heart Transplantation: A Validation of the Danish Comorbidity Index for Acute Myocardial Infarction, Charlson Comorbidity Index, and Elixhauser Comorbidity Index. Transplant Direct 2023, 9(4):e1438. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Epstein RH, Dexter F: Development and validation of a structured query language implementation of the Elixhauser comorbidity index. J Am Med Inform Assoc 2017, 24(4):845–850. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 15. Mehta HB, Sura SD, Adhikari D, Andersen CR, Williams SB, Senagore AJ, Kuo YF, Goodwin JS: Adapting the Elixhauser comorbidity index for cancer patients. Cancer 2018, 124(9):2018–2025. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Tsai J, Kim Y: Performance of the Elixhauser Comorbidity Index in Predicting Mortality Among a National US Sample of Hospitalized Homeless Adults. Med Care 2024, 62(8):543–548. [ DOI ] [ PubMed ] [ Google Scholar ] 17. Bowen M, Marwick S, Marshall T, Saunders K, Burwood S, Yahyouche A, Stewart D, Paudyal V: Multimorbidity and emergency department visits by a homeless population: a database study in specialist general practice. Br J Gen Pract 2019, 69(685):e515–e525. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Queen AB, Lowrie R, Richardson J, Williamson AE: Multimorbidity, disadvantage, and patient engagement within a specialist homeless health service in the UK: an in-depth study of general practice data. BJGP Open 2017, 1(3):bjgpopen17X100941. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 19. Blumberg MJ, Petersson AM, Jones PW, Jones AA, Panenka WJ, Leonova O, Vila-Rodriguez F, Lang DJ, Barr AM, MacEwan GW et al. : Differential sensitivity of intraindividual variability dispersion and global cognition in the prediction of functional outcomes and mortality in precariously housed and homeless adults. Clin Neuropsychol 2024:1–24. [ DOI ] [ PubMed ] [ Google Scholar ] 20. Brown RT, Evans JL, Valle K, Guzman D, Chen YH, Kushel MB: Factors Associated With Mortality Among Homeless Older Adults in California: The HOPE HOME Study. JAMA Intern Med 2022, 182(10):1052–1060. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Loubiere S, Hafrad I, Monfardini E, Mosnier M, Bosetti T, Auquier P, Mosnier E, Tinland A: Morbidity and mortality in a prospective cohort of people who were homeless during the COVID-19 pandemic. Front Public Health 2023, 11:1233020. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. Nicholas W, Greenwell L, Henwood BF, Simon P: Using Point-in-Time Homeless Counts to Monitor Mortality Trends Among People Experiencing Homelessness in Los Angeles County, California, 2015‒2019. Am J Public Health 2021, 111(12):2212–2222. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. O’Connell JJ: Zones of Excess Mortality for Homeless Adults in the US-A Half Century Later. JAMA Intern Med 2022, 182(10):1061–1062. [ DOI ] [ PubMed ] [ Google Scholar ] 24. Tinland A, Loubiere S, Cantiello M, Boucekine M, Girard V, Taylor O, Auquier P: Mortality in homeless people enrolled in the French housing first randomized controlled trial: a secondary outcome analysis of predictors and causes of death. BMC Public Health 2021, 21(1):1294. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 25. Woodman L, Staples L, Karin E, Solterbeck K, Burns N, Mitchell R, Nielssen O: Rates and causes of mortality among the homeless in Sydney. Australas Psychiatry 2023, 31(4):469–474. [ DOI ] [ PubMed ] [ Google Scholar ] 26. Zordan R, Mackelprang JL, Hutton J, Moore G, Sundararajan V: Premature mortality 16 years after emergency department presentation among homeless and at risk of homelessness adults: a retrospective longitudinal cohort study. Int J Epidemiol 2023, 52(2):501–511. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Gicas KM, Jones AA, Thornton AE, Petersson A, Livingston E, Waclawik K, Panenka WJ, Barr AM, Lang DJ, Vila-Rodriguez F et al. : Cognitive decline and mortality in a community-based sample of homeless and precariously housed adults: 9-year prospective study. BJPsych Open 2020, 6(2):e21. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 28. Gundlapalli AV, Redd A, Bolton D, Vanneman ME, Carter ME, Johnson E, Samore MH, Fargo JD, O’Toole TP: Patient-aligned Care Team Engagement to Connect Veterans Experiencing Homelessness With Appropriate Health Care. Med Care 2017, 55 Suppl 9 Suppl 2:S104–s110. [ DOI ] [ PubMed ] [ Google Scholar ] 29. Leung LB, Zhang E, Chu K, Yoo C, Gabrielian S, Der-Martirosian C: Characteristics of Veterans Experiencing Homelessness using Telehealth for Primary Care Before and After COVID-19 Pandemic Onset. J Gen Intern Med 2024, 39(Suppl 1):53–59. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 30. Nescott E, Metraux S, McDuffie MJ, Brown E: Health & Housing in Delaware:: Matching Medicaid Claims and Encounters and the Community Management Information System Databases. Dela J Public Health 2023, 9(2):18–22. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Pribish A, Khalil N, Mhaskar R, Woodard L, Mirza AS: Chronic Disease Burden of the Homeless: A Descriptive Study of Student-Run Free Clinics in Tampa, Florida. J Community Health 2019, 44(2):249–255. [ DOI ] [ PubMed ] [ Google Scholar ] 32. Smith-Bernardin S, Carrico A, Max W, Chapman S: Utilization of a Sobering Center for Acute Alcohol Intoxication. Acad Emerg Med 2017, 24(9):1060–1071. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 33. Maron SZ, Neifert SN, Ranson WA, Nistal DA, Rothrock RJ, Cooke P, Lamb CD, Cho SK, Caridi JM: Elixhauser Comorbidity Measure is Superior to Charlson Comorbidity Index In-Predicting Hospital Complications Following Elective Posterior Cervical Decompression and Fusion. World Neurosurg 2020, 138:e26–e34. [ DOI ] [ PubMed ] [ Google Scholar ] 34. Zhang F, Chiu Y, Ensor J, Mohamed MO, Peat G, Mamas MA: Elixhauser outperformed Charlson comorbidity index in prognostic value after ACS: insights from a national registry. J Clin Epidemiol 2022, 141:26–35. [ DOI ] [ PubMed ] [ Google Scholar ] 35. Ferguson JM, Wray CM, Jacobs J, Greene L, Wagner TH, Odden MC, Freese J, Van Campen J, Asch SM, Heyworth L et al. : Variation in initial and continued use of primary, mental health, and specialty video care among Veterans. Health Serv Res 2023, 58(2):402–414. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. Jones AL, Gordon AJ, Gabrielian SE, Montgomery AE, Blosnich JR, Varley AL, deRussy AJ, Austin EL, Hoge AE, Kim YI et al. : Perceptions of Care Coordination Among Homeless Veterans Receiving Medical Care in the Veterans Health Administration and Community Care Settings: Results From a National Survey. Med Care 2021, 59(6):504–512. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 37. Swankoski KE, Reddy A, Grembowski D, Chang ET, Wong ES: Intensive care management for high-risk veterans in a patient-centered medical home - do some veterans benefit more than others? Healthc (Amst) 2023, 11(2):100677. [ DOI ] [ PubMed ] [ Google Scholar ] 38. Laliberte AZ, Salvi A, Hooker E, Roth B, Handley R, Carlson K, Hynes D, Tuepker A, Chen JI: Factors associated with a lack of health care utilization among Veterans after a positive suicide screen in the emergency department. Acad Emerg Med 2023, 30(4):321–330. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 39. Holliday R, Smith AA, Kinney AR, Forster JE, Bahraini N, Monteith LL, Brenner LA: Research Letter: Traumatic Brain Injury Among Veterans Accessing VA Justice-Related Services. J Head Trauma Rehabil 2023, 38(2):184–190. [ DOI ] [ PubMed ] [ Google Scholar ] 40. Tsai J, Szymkowiak D, Jutkowitz E: Developing an operational definition of housing instability and homelessness in Veterans Health Administration’s medical records. PLoS One 2022, 17(12):e0279973. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 41. [ https://hcup-us.ahrq.gov/toolssoftware/comorbidityicd10/comorbidity_icd10.jsp ] [ Google Scholar ] 42. McLernon DJ, Giardiello D, Van Calster B, Wynants L, van Geloven N, van Smeden M, Therneau T, Steyerberg EW, topic g, Initiative otS: Assessing Performance and Clinical Usefulness in Prediction Models With Survival Outcomes: Practical Guidance for Cox Proportional Hazards Models. Ann Intern Med 2023, 176(1):105–114. [ DOI ] [ PubMed ] [ Google Scholar ] 43. Rahman MS, Ambler G, Choodari-Oskooei B, Omar RZ: Review and evaluation of performance measures for survival prediction models in external validation settings. BMC Med Res Methodol 2017, 17(1):60. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. Sonaglioni A, Lombardo M, Albini A, Noonan DM, Re M, Cassandro R, Elia D, Caminati A, Nicolosi GL, Harari S: Charlson comorbidity index, neutrophil-to-lymphocyte ratio and undertreatment with renin-angiotensin-aldosterone system inhibitors predict in-hospital mortality of hospitalized COVID-19 patients during the omicron dominant period. Front Immunol 2022, 13:958418. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 45. Cho SI, Yoon S, Lee HJ: Impact of comorbidity burden on mortality in patients with COVID-19 using the Korean health insurance database. Sci Rep 2021, 11(1):6375. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 46. Kim DH, Park HC, Cho A, Kim J, Yun KS, Kim J, Lee YK: Age-adjusted Charlson comorbidity index score is the best predictor for severe clinical outcome in the hospitalized patients with COVID-19 infection. Medicine (Baltimore) 2021, 100(18):e25900. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 47. Jones CD, Falvey J, Hess E, Levy CR, Nuccio E, Barón AE, Masoudi FA, Stevens-Lapsley J: Predicting Hospital Readmissions from Home Healthcare in Medicare Beneficiaries. J Am Geriatr Soc 2019, 67(12):2505–2510. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 48. Hum A, George PP, Tay RY, Wu HY, Ali NB, Leong I, Chin JJ, Lee A, Tan L, Koh M: Prognostication in Home-Dwelling Patients with Advanced Dementia: The Palliative Support DEMentia Model (PalS-DEM). J Am Med Dir Assoc 2021, 22(2):312–319.e313. [ DOI ] [ PubMed ] [ Google Scholar ] 49. Stensrud MJ, Hernan MA: Why Test for Proportional Hazards? JAMA 2020, 323(14):1401–1402. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 50. Demidenko E: The p-Value You Can’t Buy. Am Stat 2016, 70(1):33–38. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials 1 NIHMS2071734-supplement-1.docx (85.6KB, docx) Data Availability Statement The data that support the findings of this study are available from the U.S. Department of Veterans Affairs. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the authors with the permission of the U.S. Department of Veterans Affairs. ACTIONS View on publisher site PDF (403.5 KB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Bacterial Overgrowth Associated With Chronic Multisymptom Illness Complex#829387
• Trends in cardiovascular mortality (1999-2023) in the United States: an analysis of underlying causes of death using the CDC WONDER database.#375543
• Understanding Mortality Data: A Step-by-Step Guide to CDC WONDER, Joinpoint Analysis, and Forecasting Models.#182415
• Trauma Specific Frailty Index#825978
• Cardiovascular Comorbidities and Bronchiectasis#1032841