Factors independently associated with in-hospital HRQoL in adults treated at a high-complexity institution in a middle-income country

Abstract

Introduction

Health systems are increasingly oriented toward value-based care, emphasizing patient-centered outcomes such as health-related quality of life (HRQoL). During hospitalization, HRQoL reflects the interplay of sociodemographic, geographic, and clinical factors, which may be particularly influential in middle-income settings where structural inequities can affect patient-reported outcomes.

Objective

To identify factors independently associated with HRQoL measured during hospitalization in adults treated at a high-complexity institution in a middle-income country.

Methods

We conducted a cross-sectional analytical study nested within an institutional prospective cohort (2023–2025). We included hospitalized adults with a complete in-hospital HRQoL assessment using the EQ-5D-3L descriptive system and the EQ visual analogue scale (EQ VAS). The EQ VAS score was analyzed as the primary continuous outcome. Associations were evaluated using bivariate analyses and multivariable linear regression.

Results

A total of 46,083 patients were included, with a median EQ-VAS score of 70 (IQR, 60–80). Functional impairment was most commonly reported in usual activities (62.46%). In the multivariable linear regression model, lower EQ-VAS scores were independently associated with increasing age (β = − 0.150 per year) and ICU admission (β = − 1.661), whereas professional education (β = 2.364) and surgical admission (β = 1.672) were associated with higher scores. Across diagnostic groups, mental and behavioural disorders (β = − 2.949) and respiratory diseases (β = − 2.036) showed the strongest negative associations, while pregnancy-related conditions were associated with the highest scores (β = 3.982).

Conclusion

EQ VAS scores measured during hospitalization were independently associated with sociodemographic and clinical factors, underscoring the multifactorial determinants of self-rated health in acute hospital settings.

Introduction

The transition of health systems toward value-based care has redefined clinical success beyond traditional outcomes such as survival or resolution of acute disease, emphasizing functional recovery and patients’ perceived well-being [1, 2]. Within this paradigm, health-related quality of life (HRQoL) has emerged as a key patient-centered outcome for evaluating healthcare effectiveness, particularly in high-complexity institutions where disease severity and therapeutic intensity are substantial [3, 4].

During hospitalization, HRQoL reflects the interplay of sociodemographic, geographic, and clinical factors, which may be particularly influential in middle-income settings where structural inequities can affect patient-reported outcomes These interactions may be especially relevant in middle-income countries, where structural inequalities, heterogeneous access to care, and geographic disparities can shape patients’ experiences and reported health status [5]. Previous studies have demonstrated that sociodemographic characteristics, socioeconomic position, and contextual factors are associated with variations in HRQoL among hospitalized individuals, reflecting differences in vulnerability, coping capacity, and perceived health [6–8].

Despite increasing use of patient-reported outcome measures, most research has focused on post-discharge trajectories or chronic disease management, while the determinants of HRQoL during the acute hospitalization phase remain insufficiently explored. Hospitalization itself represents a period of dynamic clinical instability, during which disease severity, treatment intensity, and functional limitations may transiently influence perceived health status. Instruments such as the EQ-5D and its visual analogue scale provide sensitive measures capable of capturing these changes in real time [9, 10].

Routinely collected clinical and contextual data obtained during hospitalization may serve as proxies of physiological stability, healthcare delivery processes, and social vulnerability, potentially shaping patients’ perceptions of health while receiving inpatient care [11, 12]. However, the independent contribution of these factors to HRQoL assessed during hospitalization has not been fully characterized, particularly in high-complexity referral centers within middle-income settings.

Understanding these relationships is essential for informing patient-centered interventions, optimizing clinical decision-making, and advancing value-based healthcare strategies. Therefore, this study aimed to identify factors independently associated with HRQoL measured during hospitalization in adults treated at a high-complexity institution in a middle-income country.

Methods

Study design and population

This cross-sectional analytical study was nested within an ongoing institutional prospective follow-up cohort established in 2023 at the Hospital Internacional de Colombia–Fundación Cardiovascular de Colombia (HIC–FCV), a high-complexity tertiary referral center located in a middle-income country. The parent cohort prospectively collects standardized clinical and patient-reported information as part of routine hospital care.

This analysis included hospitalizations from January 2023 to December 2025, focusing on Health-Related Quality of Life (HRQoL) assessed during the index stay. HRQoL was evaluated once, typically within 48 h of admission or upon clinical stability, ensuring the patient could provide a reliable self-report. Hospitalized adult patients are systematically enrolled using REDCap electronic data capture tools [13, 14]. Trained personnel administer structured HRQoL assessments and document sociodemographic and clinical information obtained from electronic health records. Although the institutional program includes post-discharge follow-up at 1, 6, and 12 months, this study was restricted to in-hospital measurements to preserve the cross-sectional analytical design.

Eligible participants were adults aged ≥18 years admitted to medical or surgical services with hospital stays longer than 24 h and a completed in-hospital HRQoL assessment. Patients were excluded if essential clinical data were missing, if transfer occurred before completion of hospitalization, or if psychiatric, cognitive, or clinical conditions prevented reliable self-reporting. No additional study-specific exclusions were applied. The final analytical sample included only patients with complete HRQoL assessment during hospitalization. The participant selection process is shown in Fig. 1.

Fig. 1.

Study population selection and derivation of the final analytical cohort. Flow diagram illustrating the identification, eligibility assessment, and inclusion of hospitalized adults from the institutional follow-up program database between 2023 and 2025. Subsequent rehospitalizations of patients with more than one admission during the study period were excluded, along with international patients and records with incomplete EQ-5D-3L domain responses. The final analytical sample comprised 46,083 unique patients, each contributing a single index hospitalization with a complete in-hospital EQ-5D-3L and EQ VAS assessment

Data management and information engineering

A structured data engineering framework was implemented to ensure data integrity, traceability, and analytical reproducibility. Data were automatically extracted from institutional information systems using standardized query algorithms, followed by validation, harmonization, and normalization procedures prior to integration into REDCap, which served as the centralized repository for data auditing and preparation of the analytical dataset.

Quality control procedures included automated range validation and controlled dataset versioning. The proportion of missing data was evaluated across all variables, showing an overall level of missingness below 5% with no evidence of systematic patterns.

Outcome measurement

The primary outcome was health-related quality of life (HRQoL) assessed during hospitalization using the EuroQol EQ-5D-3L instrument [15], a validated generic patient-reported outcome measure that evaluates mobility, self-care, usual activities, pain or discomfort, and anxiety or depression across three levels of severity, (no problems:level 1, some/moderate problems:level 2, and extreme problems:level 3). Participants additionally rated their overall health using the EuroQol Visual Analogue Scale (EQ-VAS), ranging from 0 (worst imaginable health) to 100 (best imaginable health).

Although the EQ-5D descriptive system allows calculation of a health utility index based on country-specific value sets, no nationally validated EQ-5D-3L tariff is currently available for Colombia. Consequently, calculation of a population-specific utility index was not feasible. To avoid applying value sets derived from other countries that may not reflect local health preferences, the EQ-VAS score was selected as the primary outcome measure. The EQ-VAS was analyzed as a continuous dependent variable representing global perceived well-being during hospitalization and providing a direct, preference-independent measure of patients’ self-rated health within the Colombian inpatient context.

Independent variables

Independent variables were defined a priori based on conceptual relevance and existing evidence regarding factors associated with patient-reported outcomes and corresponded to those presented in the analytical tables. Sociodemographic characteristics included age, sex, marital status, educational attainment, insurance regime, area of residence and year of hospitalization. In Colombia, the contributory regime covers formally employed, self-employed, and retired individuals through mandatory income-based contributions, whereas the subsidized regime provides publicly funded coverage for people without sufficient ability to pay. The special regime comprises separate healthcare arrangements established for specific occupational groups, including members of the Armed Forces and National Police, public-school teachers, Ecopetrol employees, and certain public university employees. Clinical and care-related variables comprised type of admission categorized as medical or surgical, total length of hospital stay, admission to the intensive care unit at any time during the index hospitalization, systolic and diastolic blood pressure recorded during hospitalization, and vital status at discharge. Because total length of stay, intensive care unit admission, and vital status at discharge may have occurred or become fully known after the in-hospital EQ-VAS assessment, these variables were included as retrospective proxy indicators of clinical severity and the overall hospital course, rather than as temporally antecedent or causal determinants of the EQ-VAS score. Accordingly, their coefficients should be interpreted as adjusted associations and not as evidence of directionality or causation. Primary discharge diagnoses were coded according to the International Classification of Diseases, Tenth Revision (ICD-10) [16], and grouped into major diagnostic chapters to approximate clinical case mix across hospitalized patients.

Statistical analysis

Descriptive analyses were performed to characterize the study population and HRQoL outcomes. Continuous variables were assessed for normality using the Shapiro–Wilk test and summarized as means with standard deviations or medians with interquartile ranges, as appropriate, whereas categorical variables were presented as frequencies and percentages. Bivariate associations between independent variables and EQ-VAS scores were evaluated using the Mann–Whitney U test for two-group comparisons and the Kruskal–Wallis test for comparisons involving more than two groups. Correlations between continuous variables were examined using Spearman’s rank correlation coefficient.

To identify factors independently associated with EQ-VAS scores, multivariable linear regression models were fitted using robust standard errors. An initial fully specified model included all prespecified sociodemographic, clinical, care-related, and contextual variables, irrespective of their bivariate statistical significance. A purposeful sequential model-reduction approach was subsequently applied to obtain a more parsimonious final model. Variables were removed one at a time when their exclusion did not materially alter the direction or magnitude of the remaining coefficients or overall model performance. Sex, year of discharge, and systolic blood pressure were excluded during this process, whereas variables considered clinically and conceptually relevant, including diagnostic case mix, were retained. Diagnostic case mix was incorporated through grouped ICD-10 categories.

Model performance was evaluated using coefficients of determination (R² and adjusted R²), and results are presented as regression coefficients with corresponding 95% confidence intervals. Sensitivity analyses were conducted to evaluate the robustness of findings by re-estimating models after excluding patients who died during hospitalization, fitting median regression models with bootstrap standard errors to address potential non-normality of EQ-VAS scores, and sequentially adjusting models for markers of clinical severity including length of stay, intensive care unit admission, and hemodynamic parameters. A two-sided p value below 0.05 was considered statistically significant. All analyses were performed using Stata version 16 (StataCorp, College Station, TX, USA).

Ethical considerations

The study was conducted in accordance with the Declaration of Helsinki and national regulations governing research involving human participants. The protocol was approved by the Institutional Ethics Committee (approval number CEI-2023-06526). All participants provided written informed consent prior to inclusion, and data were analyzed in anonymized form to ensure confidentiality and data protection.

Results

A total of 46,083 hospitalized adults were included in the analysis. The median EQ-VAS score during hospitalization was 70 (IQR 60–80), reflecting overall preserved but heterogeneous perceived health status. Medical admissions accounted for 59.88% (n = 27,591) of hospitalizations, whereas surgical patients represented 40.12% (n = 18,492). The median EQ VAS score was 70 in each discharge year. However, the overall distribution differed significantly across years (p < 0.001), with the 75th percentile decreasing from 80 in 2023 and 2024 to 70 in 2025. Sex distribution was balanced (49.05% women and 50.95% men). In contrast, social characteristics were associated with HRQoL variation. Lower educational attainment was associated with progressively reduced EQ-VAS values, particularly among individuals without formal education (median 68 [57–71]). Age demonstrated an inverse correlation with EQ-VAS (ρ = −0.286; p < 0.001). Length of stay also correlated negatively with perceived health (ρ = −0.120; p < 0.001), while diastolic blood pressure showed a weak positive correlation (ρ = 0.067; p < 0.001) (Table 1).

Table 1.

Distribution of EQ-VAS scores according to clinical and sociodemographic characteristics in hospitalized adults

Variable	Category	n (%)	p25	p50	p75	P value
Sex	Female	22,603 (49.05)	60	70	80	< 0.001
	Male	23,480 (50.95)	60	70	78
Area of residence	Rural	6,719 (14.58)	60	70	80	0.539
	Urban	39,364 (85.42)	60	70	80
Insurance type	Contributory	18,555 (40.27)	60	70	80	< 0.001
	Subsidized	18,750 (40.68)	60	70	76
	Special	8,778 (19.05)	61	70	80
Marital status	With partner	24,822 (53.86)	60	70	80	< 0.001
	Without partner	16,949 (36.78)	60	70	79
	Not declared	4,312 (9.36)	60	70	80
Education level	None	2,594 (5.63)	57	68	71
	Primary	17,397 (37.75)	60	70	75	< 0.001
	Secondary	16,042 (34.81)	60	70	80
	Technical	2,696 (5.85)	67	70	80
	Professional	5,693 (12.36)	67	70	80
	Missing	1,661 (3.60)	60	70	74
Surgical condition	Medical	27,591 (59.88)	60	70	75	< 0.001
	Surgical	18,492 (40.12)	62	70	80
Vital status	Deceased	1,326 (2.88)	50	60	70	< 0.001
	Alive	44,757 (97.12)	60	70	80
ICU	Yes	14,272 (30.97)	60	70	71	< 0.001
	No	31,811 (69.03)	60	70	80
Discharge year	2023	17,287 (37.51)	61	70	80	< 0.001
	2024	16,524 (35.85)	60	70	80
	2025	12,272 (26.64)	60	70	70
			ρ
Age, years	Median (IQR)	60 (43–73)	− 0.286			<0.001
LOS, days	Median (IQR)	4 (2–9)	− 0.120			<0.001
SBP, mmHg	Median (IQR)	127 (113–144)	− 0.028			<0.001
DBP, mmHg	Median (IQR)	75 (66–85)	0.067			<0.001

Values are presented as n (%) for categorical variables and as medians with interquartile ranges (IQR) for continuous variables. EQ-VAS scores are expressed on a 0–100 scale, where higher values indicate better perceived health status. Percentiles are reported as p25 (25th percentile), p50 (median), and p75 (75th percentile). Group comparisons were performed using the Mann–Whitney U test for dichotomous variables and the Kruskal–Wallis test for variables with more than two categories. Associations between continuous variables and EQ-VAS were evaluated using Spearman’s rank correlation coefficient (ρ). Statistical significance was defined as a two-sided p value <0.05. EQ-VAS, EuroQol Visual Analogue Scale; IQR, LOS, length of stay (days); SBP, systolic blood pressure (mmHg); DBP, diastolic blood pressure (mmHg); ICU, intensive care unit

EQ-VAS scores differed significantly across ICD-10 diagnostic groups (p < 0.001). Circulatory diseases were the most frequent (33.55%; median 70 [60–71]), followed by neoplasms (11.85%; 70 [60–80]), genitourinary (8.67%; 70 [62–80]), digestive (7.88%; 70 [61–80]), and factors influencing health status (7.24%; 70 [60–80]); pregnancy and puerperium showed the highest median (2.78%; 80 [70–80]), whereas respiratory (3.55%) and mental disorders (0.60%) had the lowest medians (65 [50–70] and 69 [51–74], respectively), with remaining categories each representing ≤ 4.54% of patients (Table 2).

Table 2.

Health-Related Quality of Life (EQ-VAS) According to ICD-10 Diagnostic Groups During Hospitalization

ICD-10 Group	n (%)	p25	p50	p75	P value
Diseases of the circulatory system (I00–I99)	15,459 (33.55)	60	70	71	< 0.001
Neoplasms (C00–D48)	5,462 (11.85)	60	70	80
Diseases of the genitourinary system (N00–N99)	3,997 (8.67)	62	70	80
Diseases of the digestive system (K00–K93)	3,629 (7.88)	61	70	80
Factors influencing health status and contact with health services (Z00–Z99)	3,337 (7.24)	60	70	80
Diseases of the musculoskeletal system and connective tissue (M00–M99)	2,091 (4.54)	63	70	80
Injury, poisoning and certain other consequences of external causes (S00–T98)	2,054 (4.46)	61	70	80
Diseases of the respiratory system (J00–J99)	1,635 (3.55)	50	65	70
Certain infectious and parasitic diseases (A00–B99)	1,452 (3.15)	60	70	80
Pregnancy, childbirth and the puerperium (O00–O99)	1,280 (2.78)	70	80	80
Diseases of the nervous system (G00–G99)	1,231 (2.67)	60	70	78
Endocrine, nutritional and metabolic diseases (E00–E90)	1,230 (2.67)	64	70	80
Symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified (R00–R99)	1,152 (2.50)	60	70	80
Diseases of the skin and subcutaneous tissue (L00–L99)	890 (1.93)	60	70	80
Diseases of the blood and blood-forming organs and certain disorders involving the immune mechanism (D50–D89)	531 (1.15)	60	70	76
Mental and behavioural disorders (F00–F99)	278 (0.60)	51	69	74
Congenital malformations, deformations and chromosomal abnormalities (Q00–Q99)	129 (0.28)	70	75	80
Diseases of the ear and mastoid process (H60–H95)	114 (0.25)	65	70	80
External causes of morbidity and mortality (V01–Y98)	37 (0.08)	60	70	80
Diseases of the eye and adnexa (H00–H59)	36 (0.08)	56.5	69.5	79.5
Codes for special purposes (U00–U99)	33 (0.07)	60	70	78
Certain conditions originating in the perinatal period (P00–P96)	26 (0.06)	70	77.5	80

Values are presented as number (percentage) for categorical variables and median with interquartile range (p25–p75) for EQ-VAS scores. The EQ-VAS ranges from 0 (worst imaginable health state) to 100 (best imaginable health state). P values were derived from Kruskal–Wallis tests comparing EQ-VAS distributions across ICD-10 diagnostic groups. ICD-10 categories correspond to major chapters of the International Classification of Diseases, Tenth Revision

Health status measured by EQ-5D showed predominantly mild functional impairment. Absence of mobility limitations was reported by 47.96%, while 40.44% had some difficulty and 11.61% were bedridden. Similar patterns were observed for self-care (47.50% without limitations; 43.76% with some difficulty; 8.74% unable to perform independently). Limitations were more pronounced in usual activities, where 53.15% reported some problems and 9.31% were unable to perform them. Pain or discomfort was absent in 55.95%, moderate in 40.02%, and severe in 4.03%. Most patients (70.35%) reported no anxiety or depression, whereas 27.46% and 2.19% experienced moderate and severe symptoms, respectively (Fig. 2).

Fig. 2.

Distribution of EQ-5D Dimensions during hospitalization. Bar charts display the proportion of patients reporting Level 1 (no problems), Level 2 (some/moderate problems), or Level 3 (extreme problems) across the five EQ-5D-3L domains (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression). The response levels are identified in the visual legend included in the figure. The box plot illustrates the distribution of EQ-VAS scores, including the median, interquartile range, and extreme values, in the overall hospitalized population (n = 46,083)

In the multivariable linear regression analysis, several sociodemographic and clinical factors were independently associated with EQ-VAS scores. Increasing age was significantly related to poorer perceived health status, with a decrease of 0.15 points for each additional year (β = − 0.150; 95% CI − 0.158 to − 0.143; p < 0.001). Compared with individuals living with a partner, those without a partner or with unreported marital status had lower scores, whereas higher educational attainment showed a gradual positive association with EQ-VAS values. Residence in urban areas and enrollment in the subsidized insurance regime were associated with lower scores, while the special insurance regime was linked to better perceived health. Several diagnostic categories demonstrated significant associations, notably lower scores in mental disorders and respiratory diseases, and higher scores in pregnancy-related conditions, congenital anomalies, and ear disorders. From a care-related perspective, surgical admission and higher diastolic blood pressure values were positively associated with EQ-VAS, whereas longer hospital stay and intensive care unit hospitalization were negatively associated. The final multivariable model explained a modest proportion of the variance in perceived health status (R² = 0.123) (Table 3).

Table 3.

Multivariable linear regression analysis of factors associated with EQ-VAS scores during hospitalization

Variable	β coefficient	95% CI	p-value
Age (years)	− 0.150	− 0.158 to − 0.143	< 0.001
Marital status
No partner	− 0.771	− 1.027 to − 0.514	< 0.001
Not reported	− 0.696	− 1.141 to − 0.252	0.002
Education level
Primary	0.960	0.415 to 1.505	0.001
Secondary	1.364	0.795 to 1.933	< 0.001
Technical/technologist	2.323	1.591 to 3.054	< 0.001
Professional	2.364	1.720 to 3.007	< 0.001
No information	− 0.189	− 1.010 to 0.633	0.652
Area of residence
Urban	− 0.461	− 0.799 to − 0.123	0.008
Insurance regime
Subsidized	− 0.756	− 1.040 to − 0.472	< 0.001
Special regime	1.022	0.653 to 1.391	< 0.001
Diagnosis group (ICD-10 chapters)
Neoplasms	1.016	0.211 to 1.822	0.013
Blood/immune disorders	− 0.265	− 1.631 to 1.101	0.704
Endocrine/metabolic disorders	1.271	0.249 to 2.293	0.015
Mental disorders	− 2.949	− 4.711 to − 1.188	0.001
Nervous system diseases	− 1.136	− 2.205 to − 0.068	0.037
Eye disorders	− 4.016	− 9.598 to 1.565	0.158
Ear disorders	4.288	1.867 to 6.709	0.001
Circulatory system diseases	0.622	− 0.164 to 1.408	0.121
Respiratory system diseases	− 2.036	− 3.020 to − 1.052	< 0.001
Digestive system diseases	0.889	0.030 to 1.747	0.043
Skin disorders	− 0.163	− 1.282 to 0.957	0.776
Musculoskeletal diseases	0.557	− 0.382 to 1.495	0.245
Genitourinary diseases	1.379	0.514 to 2.244	0.002
Pregnancy/childbirth	3.982	3.033 to 4.932	< 0.001
Perinatal conditions	1.388	− 0.558 to 3.334	0.162
Congenital anomalies	3.710	1.540 to 5.881	0.001
Symptoms/signs	1.324	0.189 to 2.458	0.022
Injury/poisoning	− 0.697	− 1.633 to 0.239	0.144
Special codes	3.378	− 1.381 to 8.137	0.164
External causes	− 3.530	− 7.387 to 0.328	0.073
Health services factors	− 0.999	− 1.866 to − 0.131	0.024
Admission condition
Surgical	1.672	1.272 to 2.072	< 0.001
Alive at discharge	8.549	7.533 to 9.565	< 0.001
Length of stay (days)	− 0.098	− 0.170 to − 0.026	0.008
Diastolic blood pressure (mmHg)	0.028	0.020 to 0.037	< 0.001
ICU stay	− 1.661	− 2.182 to − 1.140	< 0.001

β coefficients represent adjusted mean differences in EQ-VAS score (0–100 scale) associated with each independent variable. Positive coefficients indicate higher perceived health status, whereas negative coefficients indicate lower perceived health status. The model was adjusted for sociodemographic characteristics, insurance regime, ICD-10 diagnostic groups, admission condition, length of stay, diastolic blood pressure, and ICU stay. Statistical significance was defined as a two-sided p value < 0.05. EQ-VAS: EuroQol Visual Analogue Scale; ICD-10: International Classification of Diseases, Tenth Revision; ICU: Intensive Care Unit; CI confidence interval. The table presents all coefficients, including statistically non-significant associations, for the variables retained in the final model

The model did not show evidence of relevant multicollinearity, with a mean variance inflation factor (VIF) of 2.10, a value that falls within the acceptable range for multivariable linear regression. Plots assessing model assumptions: Fig. 3a Q-Q plot of standardized residuals against the theoretical normal distribution, showing an overall approximation to normality with some deviations in the tails; Fig. 3b histogram of residuals with an overlaid density curve, indicating an approximately symmetric distribution centered around zero. Overall, these findings suggest an adequate global model fit, with the presence of potential extreme values at the distribution limits.

Fig. 3.

Diagnostic assessment of linear regression model assumptions. Panel (a) shows the normal probability (Q–Q) plot of standardized residuals against the theoretical normal distribution, illustrating overall alignment with the normality assumption with mild deviations in the tails. Panel (b) displays the histogram and kernel density estimate of residuals, indicating an approximately symmetric distribution centered around zero

Sensitivity analyses

The results of sensitivity analyses were consistent with the main findings. After excluding patients who died during hospitalization, the direction and magnitude of regression coefficients remained largely unchanged, indicating that the associations observed were not driven solely by end-of-life clinical status. Median regression models with bootstrap standard errors yielded effect estimates similar to those obtained in the primary models, confirming the robustness of results to distributional assumptions of the outcome variable. Furthermore, sequential adjustment for clinical severity indicators, including length of stay, intensive care unit admission, and blood pressure parameters, did not materially modify the associations between sociodemographic factors and health-related quality of life. Overall, these analyses support the stability and reliability of the main results.

Discussion

In this large cohort of 46,083 hospitalized adults treated at a high-complexity institution in a middle-income country, multiple sociodemographic and clinical factors were independently associated with health-related quality of life measured using the EQ-VAS. The consistency of associations across sensitivity analyses supports the robustness of these findings and highlights the multidimensional determinants of perceived health during hospitalization.

Increasing age showed a consistent inverse association with EQ VAS scores. This finding is consistent with longitudinal studies reporting poorer health-related quality of life among older adults following hospitalization, which may reflect reduced functional reserve and a greater burden of multimorbidity [17–19].

From a socioeconomic perspective, lower educational attainment, affiliation with the subsidized insurance regime, and urban residence were associated with lower EQ-VAS scores, reinforcing the well-established role of social determinants in shaping perceived health status [20, 21]. In the Colombian context, where structural inequities in access to care and living conditions remain pronounced, these results suggest that social vulnerability continues to influence health perception even within high-complexity hospital environments [22].

Diagnostic case mix also contributed to variability in EQ VAS scores. Mental and behavioural disorders and respiratory diseases showed the strongest negative associations, consistent with evidence demonstrating the impact of psychological distress and symptom burden on patient-reported outcomes [23, 24]. Conversely, pregnancy- and puerperium-related conditions were independently associated with higher EQ VAS scores after adjustment for age and the other covariates included in the multivariable model. These findings underscore the importance of accounting for diagnostic heterogeneity when evaluating self-rated health among hospitalized populations [25, 26].

Regarding markers of clinical severity, ICU admission and prolonged hospital stay were independently associated with lower EQ VAS scores, consistent with prior studies describing the functional and psychological burden of critical illness [27, 28]. Previous studies have similarly reported poorer patient-reported outcomes among ICU survivors, supporting the association between critical illness and impaired health-related quality of life [28–30].

Survival at discharge was strongly associated with higher EQ-VAS scores. Although some methodological approaches exclude deceased patients to avoid potential bias in longitudinal analyses [31–33], we intentionally retained these individuals in this cross-sectional framework to capture the full spectrum of clinical severity during hospitalization. Excluding deceased patients could have attenuated the relationship between disease severity and perceived health status. Notably, sensitivity analyses excluding these cases yielded comparable effect estimates, indicating that the observed associations were not driven solely by terminal clinical conditions.

Similarly, ICD-10 diagnostic groups were retained in the adjusted model to account for clinical case mix. Although diagnostic categories are not direct causal determinants of health-related quality of life, their inclusion enabled better representation of clinical heterogeneity and reduced residual confounding related to underlying disease burden. Incorporating diagnostic and comorbidity information into multivariable models is a well-established epidemiological approach when analyzing patient-reported outcomes, as it captures differences in patient complexity that influence perceived health status [34, 35]. Case-mix adjustment frameworks consistently identify primary diagnosis and functional status as key contributors to variability in health outcomes across heterogeneous hospital populations, improving comparability and minimizing bias in quality-of-care evaluations [36].

The final model explained 12.3% of the variance in EQ-VAS scores (R² = 0.123). Although modest, this level of explained variance is expected for health-related quality of life outcomes, which reflect inherently multifactorial and subjective constructs influenced by biological, psychological, and social dimensions that are only partially captured by clinical variables [37–39]. Overall, these findings support the conceptualization of intrahospital HRQoL as a multidimensional outcome shaped by the interaction between clinical severity, social context, and individual patient characteristics, particularly within middle-income healthcare settings.

Limitations

When interpreting these findings, some methodological considerations should be acknowledged. The cross-sectional design with in-hospital assessment limits the ability to establish temporal relationships between the evaluated factors and health-related quality of life. Additionally, as the study was conducted in a single high-complexity center, extrapolation of the findings to other healthcare settings should be undertaken with caution. The EQ VAS is a subjective measure of self-rated overall health and has more limited interpretability than preference-based utility measures. It does not generate a health utility value, cannot be directly used to estimate quality-adjusted life-years, and the clinical meaning of differences in EQ VAS scores may vary across populations and clinical contexts. Although validated instruments were used and multivariable adjustment was performed, the potential influence of unmeasured factors associated with quality of life, such as baseline functional status or psychosocial characteristics, cannot be entirely excluded. Finally, assessment at a single time point during hospitalization does not allow evaluation of longitudinal trajectories or post-discharge changes.

Conclusion

In this study of adults hospitalized at a high-complexity institution in a middle-income country, HRQoL during hospitalization was influenced by a combination of sociodemographic and clinical factors. Older age, social vulnerability, respiratory and mental health diagnoses, prolonged hospitalization, and intensive care exposure were independently associated with poorer perceived health status. These findings highlight the multifactorial factors associated with HRQoL in acute hospital settings and emphasize the importance of integrating patient-reported outcomes into patient-centered strategies and health policy development within value-based healthcare systems..

Author contributions

ML: Conceptualization, methodology, database development and management, data analysis and interpretation, drafting of the initial manuscript, critical revision of the manuscript, review and approval of the final version, and accountability for all aspects of the work. AH: Methodology, data analysis and interpretation, drafting of the initial manuscript, critical revision of the manuscript, review and approval of the final version, and accountability for all aspects of the work. AB: Conceptualization, critical revision of the manuscript, review and approval of the final version, and accountability for all aspects of the work. GC: Data collection, critical revision of the manuscript, review and approval of the final version, and accountability for all aspects of the work. LAO: Data collection, critical revision of the manuscript, review and approval of the final version, and accountability for all aspects of the work. EFM: Methodology, data analysis and interpretation, drafting of the initial manuscript, critical revision of the manuscript, review and approval of the final version, and accountability for all aspects of the work.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

All data supporting the findings of this study are available in the institution's repository and can be provided by contacting the correspondence author.

Declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Consent for publication

Not applicable.

Informed consent

Written informed consent was obtained from all individual participants included in the study.

Research involving human participants

This study was conducted in accordance with the Declaration of Helsinki and national regulations governing research involving human participants. The protocol was approved by the Institutional Ethics Committee (approval number CEI-2023-06526).