Mortality and Practice Patterns in the MEXHEMO Hemodialysis Cohort

Abstract

Introduction

Chronic kidney disease represents a major public health challenge in Mexico. In the absence of a national hemodialysis (HD) registry, epidemiological data on patient characteristics, treatment practices, and outcomes remain scarce. The MEXHEMO project aims to characterize the clinical profile and outcomes of Mexican patients undergoing chronic HD.

Methods

This was a retrospective, observational cohort study. Centers nationwide were invited to report data from patients receiving maintenance HD between January 1 and June 30, 2023. Demographic, clinical, dialysis-related, and laboratory information was collected. Vital status was ascertained from medical records at each dialysis center, with 6 months of follow-up. Outcomes were mortality rates, standardized mortality ratios (SMRs), and risk factors for death. SMRs were calculated using the 2019 United States Renal Data System (USRDS) and the 2024 Chilean HD registry as reference populations.

Results

A total of 11,779 patients from 41 clinics were included. Mean age was 52 ± 16 years; 57% were male, and 92% had social security coverage. High-flux HD was prescribed in 86%, and 43% underwent dialyzer reuse. The crude mortality rate was 130 (95% confidence interval [CI]: 121–139) deaths/1000 patient-yrs. SMR were 1.31 (95% CI: 1.25–1.38) compared with the USRDS, and 1.57 (95% CI: 1.52–1.62) compared with the Chilean registry. Lack of social security coverage, hypoalbuminemia, hemoglobin (Hb) ≤ 9 g/dl, catheter use, and dialyzer reuse were independently associated with mortality.

Conclusion

MEXHEMO constitutes the largest cohort of HD patients reported in Mexico. The findings highlight elevated mortality despite a relatively young population, emphasizing the need for a national registry and quality improvement policies.

Graphical abstract

Chronic kidney disease is a global public health burden affecting over 800 million adults worldwide.¹ In Mexico, chronic kidney disease prevalence is estimated at 12%, and the incidence of treated end-stage kidney disease (ESKD) at 603 cases per million population.^2,3 Over the past 25 years, the prevalence of chronic HD has increased at an average annual rate of 1%, and HD has surpassed peritoneal dialysis (PD) as the predominant renal replacement therapy modality.^4,5 By the end of 2022, approximately 61,000 to 77,000 individuals were estimated to be receiving HD nationwide.5, 6, 7, 8, 9 Despite these trends, prognosis remains poor, with annual mortality rates reported up to 19%.¹⁰

Several Latin American countries, including Chile, Brazil, and Argentina, maintain national dialysis registries that routinely report outcomes among patients receiving renal replacement therapy. Mexico lacks a national registry, and available data derive largely from single-center or local cohorts,^2,11 limiting population-level assessment of survival and quality of care. Previous work from Mexico City reported that many HD patients failed to meet recommended targets: 65% had Hb < 10 g/dl, 25% had Kt/V < 1.2, 40% had serum albumin < 3.5 g/dl, and <55% met calcium and phosphorus goals.¹²

Dialyzer reuse—often driven by constrained resources and low reimbursement—is permitted by national guidelines and by the Mexican Institute of Social Security policies, which allow ≤12 reuses/dialyzer.¹³ However, national data on its prevalence and clinical implications are lacking.

To address these gaps, we analyzed data from MEXHEMO, a retrospective cohort of 11,779 chronic HD patients across Mexico. Our objectives were to describe clinical characteristics and practice patterns, estimate mortality and standardized comparisons, and identify potentially modifiable risk factors to inform quality improvement and health policy in a fragmented care environment.

Methods

Study Design and Population

MEXHEMO was a retrospective observational cohort coordinated by the Mexican Institute for Nephrology Research. In July 2023, the Mexican Institute for Nephrology Research launched an open call by sending personalized invitations to nephrologists affiliated with the institute and associated academic networks, encouraging chronic HD centers nationwide to contribute data. Participating centers retrospectively reported data on patients aged ≥ 18 years, receiving maintenance HD for ≥ 3 months between January 1 and June 30, 2023. Missing information on age or vascular access was an exclusion criterion (Figure 1). Follow-up was 6 months. Data were deidentified and extracted from medical records by trained staff at each site. The protocol was approved by the institutional review board of the National Institute of Medical Science and Nutrition “Salvador Zubirán” (Approval ID: NMM-5380-24-241).

Figure 1.

Inclusion of participants and outcomes at the end of the study period (January–July 2023). Eighty percent of invited clinics provided data, and 11.7% of subjects were excluded overall according to the prespecified inclusion and exclusion criteria. Outcome data were available for 97% of subjects at the end of the study. The sum of patients reported as dead and lost to follow-up was considered to have died during the follow up. HD, hemodialysis; PD, peritoneal dialysis.

Demographic and Clinical Information

Standardized forms were used to capture age, sex, HD initiation date, vascular access type, cause of ESKD, comorbidities (diabetes mellitus, hypertension, and cardiovascular disease), and dialysis prescription: blood flow rate, dialysate flow, ultrafiltration volume, session length, weekly frequency, and dialyzer reuse. Data were abstracted from medical records and manually entered into study forms.

Dialyzer reuse was defined as participation in a reuse program during the study period, as reported by each clinic. The number of reuses per dialyzer was not documented, and adherence to national HD norm (NOM-003-SSA3-2010)¹³ was not assessed.

Laboratory Data

Laboratory variables included Hb, ferritin, transferrin saturation, sodium, potassium, calcium, phosphorus, albumin, intact parathyroid hormone, predialysis blood urea nitrogen, and single-pool Kt/V (spKt/V). Values corresponded to the most recent result available during the study period and were obtained through each center’s routine laboratory protocols. Variables were categorized using the Dialysis Outcomes and Practice Patterns Study thresholds, except spKt/V, classified as < 1.2, 1.2–1.4, and > 1.4.¹⁴

Outcomes

Vital status and treatment status changes during follow-up were recorded. At the end of the study, patients were categorized as remaining on HD, transitioning to PD, transferring to another HD clinic, lost to follow-up, undergoing transplantation, or deceased; dates of status changes were documented. Transfers between HD centers were recorded when reported by participating clinics and typically reflected contract-related group transfers rather than individual patient relocation.

Vital status was obtained from dialysis center medical records. No external mortality registry or civil database linkage was available; therefore, only deaths reported within participating clinics during the study period were captured. Patients lost to follow-up were classified as deaths, based on the operational characteristics of outsourced dialysis in Mexico, where individual transfers between clinics are generally not feasible and disengagement from care is unlikely in the absence of a major adverse event. Uninsured patients who discontinue dialysis—often of low socioeconomic status—also face a high risk of early mortality.

Statistical Analysis

Categorical variables are presented as frequencies and percentages; continuous variables as means (SD) or medians (interquartile range), according to distribution. Crude and adjusted (model-predicted) mortality rates with 95% CI were estimated with Poisson regression using log follow-up time as offset and expressed per 1000 patient-years.

To contextualize mortality, SMRs were calculated using 2 references: the 2019 USRDS report (included in 2023 reference tables),¹⁵ and the Chilean Chronic HD Census (as of August 2024, published in January 2025).¹⁶ For USRDS comparisons, MEXHEMO rates were standardized by age, sex, cause of ESKD, and dialysis vintage to match USRDS-adjusted mortality rate, but race/ethnicity was excluded because of unavailability of data. The 2019 report was chosen to avoid pandemic-related distortions. For the Chilean census comparison, age-specific crude rates were used as reported by that registry. The SMR was computed as the ratio of projected observed to expected deaths. Given the 6-month follow-up, age-specific mortality rates from the Poisson model were annualized and applied to MEXHEMO age-group counts to project 1-year observed deaths; expected deaths were obtained by applying reference age-specific rates to the same strata.

Univariate Cox proportional hazards models were fitted for age, sex, social security coverage, dialysis vintage, diabetes mellitus, cardiovascular disease, hypertension, dialysis modality (low-flux vs. others), HD session length, sessions/wk, dialyzer reuse, ultrafiltration rate, Hb, albumin, spKt/V, sodium, potassium, calcium, phosphorus, transferrin saturation, and ferritin. A multivariable model including all covariates was fitted using a “missing data” category, with and without a reuse-by-center interaction term. As sensitivity analysis, a multivariable model was fitted using multiple imputation (30 iterations). Multivariable models were stratified by center.

All analyses were performed in Stata 17 (StataCorp LP, College Station, TX). Statistical significance was defined as P < 0.05. We followed STROBE guidance¹⁷ and used the STROBE checklist¹⁸ during editing (Supplementary Material).

Results

Characteristics of the Population

Between August and November 2023, 56 HD clinics agreed to participate, and 41 ultimately provided data. Eighteen clinics were affiliated with the study authors and contributed 54% of observations. Most centers were private, for-profit units (89%); 4 were public in-hospital clinics. Clinics in Mexico City, Querétaro, and Puebla contributed 92% of the data, and 71% of patients were treated in Mexico City (Figure 2); the remaining 8% came from Guanajuato, Jalisco, Nuevo León, and Veracruz. Of 13,347 screened patients, 11,902 met the inclusion criteria. After excluding 123 patients with missing age or vascular access data, 11,779 patients were included in the analysis (Figure 1).

Figure 2.

Geographical distribution of data from the MEXHEMO Cohort (January–July 2023). Data were obtained from 7 of 32 states of Mexico (including Mexico City). Most patients were treated in Mexico City. Darker color represents higher proportion of patients.

Baseline characteristics are summarized in Table 1. The mean age was 52 ± 16 years, and 75% were aged < 65 years. Fifty-seven percent were men. Most patients (91%) were insured by Mexican Institute of Social Security and received treatment in outsourced private facilities. Median dialysis vintage was 24 months (interquartile range: 12–48). The cause of ESKD was available for 90%: diabetes mellitus was the most frequent (38%), followed by unknown (25%), hypertension (17%), other causes (10%), glomerular disease (5%), obstructive uropathy (2%), and polycystic kidney disease (2%). Comorbidities included diabetes (53%), hypertension (59%), and cardiovascular disease (34%).

Table 1.

Baseline characteristics of patients in the MEXHEMO cohort (January 2023–July 2023)

Characteristics (total patients 11,779 | 100%)	Mean (SD) or p50 (IQR) or n (%)
Demographics
Age in yrs (n = 11,779 | 100%)
Mean	52.3 (15.9)
18–44	4061 (34%)
45–64	4723 (40%)
65–74	2061 (18%)
≥ 75	934 (8%)
Sex (n = 11,779 | 100%) male	6673 (57%)
Social Security coverage (n = 11,636 | 99%)	10,683 (92%)
Clinical characteristics
Dry weight in kg (n = 10,989 | 93%)	64.3 (14.8)
Cause of ESKD (n = 10,650 | 90%)
Diabetes mellitus	4080 (38%)
Hypertension	1831 (17%)
Glomerular disease	530 (5%)
Obstructive disease	226 (2%)
Polycystic disease (APKD)	193 (2%)
Unknown	2684 (25%)
Other	1106 (10%)
Comorbidities (n = 10,916 | 93%)
Diabetes mellitus	5914 (53%)
Hypertension	6400 (59%)
Cardiovascular disease	3682 (34%)
Dialysis parameters
Treatment modality (n = 10,804 | 92%)
Low flux HD	1320 (12%)
High flux HD	9321 (86%)
Hemodiafiltration or “expanded” HD	163 (1.5%)
Dialyzer reuse, yes (n = 11,779 | 100%)	5120 (43%)
HD session length in h (n = 11,718 | 99%)
Mean	3.1 (0.27)
< 3.0	844 (7%)
3.0 to < 3.5	9615 (82%)
3.5 to < 4.0	1008 (9%)
≥ 4.0	251 (2%)
Blood flow in ml/min (n = 11,718 | 99%)
Mean	333 (43)
< 350	7601 (65%)
350–400	2917 (23%)
> 400	1200 (10%)
Dialysate flow in ml/min (n = 11,683 | 99%)	565 (110)
Ultrafiltration rate in ml/kg/h (n = 10,805 | 92%)	10.9 (4.3)
HD frequency (n = 11,760 | 99%)
≤ 2 sessions/wk	1000 (8%)
≥ 3 sessions/wk	10.760 (92%)
spKt/V (n = 10,278 | 87%)
Mean	1.47 (0.3)
< 1.2	1098 (11%)
1.2 to < 1.4	2656 (26%)
≥ 1.4	6524 (63%)
Vascular access (n = 11,779 | 100%)
AV fistula	3477 (30%)
AV graft	27 (0.2%)
Tunneled catheter	6322 (54%)
Jugular	6047 (96%)
Subclavian	180 (3%)
Femoral	95 (1%)
Untunneled catheter	1886 (16%)
Jugular	1400 (76%)
Subclavian	117 (6%)
Femoral	369 (20%)
Other	67 (0.6%)
HD vintage in mo (n = 11,779 | 100%)
Median	24 (12 - 48)
3–6 mo	620 (5%)
6–12 mo	708 (6%)
12–36 mo	5009 (43%)
> 36 mo	5442 (46%)

APKD, adult polycystic kidney disease; AV, arteriovascular; ESKD, end-stage kidney disease; HD, hemodialysis; IQR, interquartile range; spKt/V, single-pool Kt/V.

At the end of follow-up, 10,733 patients (91%) remained on HD, 75 (0.6%) had undergone kidney transplantation, 229 (2.0%) had changed treatment status (transition to PD, transfer to another HD clinic, or palliative care), and 742 (6.3%) were classified as dead (574 deaths reported by the clinics and 168 patients lost to follow-up) (Figure 1).

Dialysis Practices and Laboratory Parameters

High-flux HD was used in 86% of patients, low-flux HD in 12%, and hemodiafiltration or medium cut-off HD in 2%. Dialyzer reuse was reported in 43%. Most patients (91%) received 3 weekly sessions lasting 3 to < 3.5 hours (mean: 3.1 hours). Mean blood flow rate was 333 ± 46 ml/min; 65% had < 350 ml/min, and 10% had ≥ 400 ml/min. Mean ultrafiltration rate was 10.9 ± 4.3 ml/kg/h. Mean spKt/V was 1.47 ± 0.3, and 89% achieved ≥ 1.2.

Vascular access included tunneled catheters in 54%, arteriovenous fistula (AVF) in 30%, and nontunneled catheters in 16%; among nontunneled catheters, 20% were femoral. Arteriovenous grafts were used in < 1%.

Laboratory results are shown in Table 2. Mean Hb was 10.0 ± 2.0 g/dl, median ferritin was 184 (interquartile range: 52–622) ng/ml, and mean transferrin saturation was 27 ± 15%. Mean albumin was 3.7 ± 0.5 g/dl. Mean potassium, calcium, and phosphorus were 5.4 ± 1.0 mEq/l, 8.6 ± 1.0 mg/dl, and 5.1 ± 2.0 mg/dl, respectively. Mean sodium and blood urea nitrogen were 138 ± 4 mEq/l and 65 ± 21 mg/dl, respectively. Missingness ranged from 4% to 18% for most variables; parathyroid hormone (median 425 pg/ml, interquartile range: 219–776) was available in 4%.

Table 2.

Laboratory data of MEXHEMO cohort and frequency by clinical category (January 2023–July 2023)

Characteristics (total patients | % of available data)	Mean (SD) or p50 (IQR) or N (%)
Hemoglobin in g/dl (n = 11,287 | 96%)
Mean	10.0 (2.0)
< 9	3440 (30%)
9 to < 10	2202 (20%)
10 to < 11	2132 (19%)
11 to < 12	1719 (15%)
12 to < 13	989 (9%)
≥ 13	805 (7%)
Ferritin in ml (n = 10,230 | 87%)
Median	184 (52–622)
< 200	5274 (52%)
200 to < 500	1901 (19%)
500 to < 800	1029 (10%)
800 to < 1200	737 (7%)
≥ 1200	1289 (12%)
Transferrin saturation in % (n = 9752 | 82%)
Mean	27 (15)
< 20	3450 (35%)
20 to < 30	3096 (32%)
30 to < 40	1711 (18%)
40 to < 50	738 (7%)
≥ 50	757 (8%)
Albumin in g/dl (n = 9941 | 85%)
Mean	3.7 (0.5)
< 3.2	1657 (17%)
3.2 to < 3.5	1321 (13%)
3.5 to < 4.0	3692 (37%)
≥ 4.0	3,271 (33%)
Calcium in mg/dl (n = 11,236 | 95%)
Mean	8.6 (1)
< 8.5	4,820 (43%)
8.5 to < 9.5	4,684 (41%)
9.5 to < 10.2	1,215 (11%)
≥ 10.2	517 (5%)
Phosphorus in mg/dl (n = 11,260 | 96%)
Mean	5.1 (2)
< 3.5	2,227 (20%)
3.5 to < 4.5	2,432 (22%)
4.5 to < 5.5	2,431 (22%)
5.5 to < 7.0	2,357 (21%)
≥ 7.0	1,813 (16%)
Predialysis BUN in mg/dl (n = 11,242 | 95%)
Mean	65 (21)
Sodium in mEq/l (n = 11,262| 96%)
Mean	138 (4)
< 130	219 (2%)
130 to < 140	6,958 (62%)
140 to < 145	3,696 (33%)
≥ 145	389 (3%)
Potassium in mEq/l (n = 11,260 | 96%)
Mean	5.4 (1)
< 3.5	117 (1%)
3.5 to < 4.0	377 (3%)
4.0 to < 5.0	3,182 (28%)
5.0 to < 5.5	2,633 (23%)
5.5 to < 6.0	2,078 (19%)
≥ 6.0	2,873 (26%)
PTH in pg/ml (n = 514 | 4%)
Median	423 (216–764)

BUN, blood urea nitrogen; IQR, interquartile range); PTH, parathyroid hormone.

Last available laboratory value. Number (n) and proportion of patients with available data are showed for each variable. Categories for each laboratory variable are based on those reported in the Dialysis Outcomes Practice Patterns Study (DOPPS) practice monitor.

Mortality Rate and SMRs

A total of 742 deaths were recorded over 5714 person-years of follow-up. The crude mortality rate was 130 per 1000 patient-yrs (95% CI: 121–139). After adjustment for age, sex, dialysis vintage, and ESKD cause, mortality was 132 per 1000 patient-yrs (95% CI: 122–142). Adjusted mortality rates by sex and age are shown in Supplementary Table S1.

Using the 2019 USRDS and the 2024 Chilean census as references, SMRs were 1.31 (95% CI: 1.25–1.38) and 1.57 (95% CI: 1.52–1.62), respectively (Figure 3), indicating higher age-group–specific mortality in MEXHEMO.

Figure 3.

Standardized mortality ratios for the MEXHEMO cohort by age group compared with reference populations. Panel a uses age-specific adjusted mortality rates from the 2019 USRDS; Panel b uses age-specific crude rates from the 2024 Chilean hemodialysis census. Points represent standardized mortality ratios and error bars 95% confidence intervals; the dashed line indicates SMR = 1.0. USRDS, United States Renal Data System.

Factors Associated With Mortality

Key risk factors (hazard ratios [HRs] and 95% CIs) are summarized in Table 3 for 3 Cox models: model 1 (univariate, Supplementary Table S2), model 2 (multivariable with missing categories, Supplementary Table S3), and model 3 (multivariable with multiple imputation, Supplementary Table S4). Full models are provided in Supplementary Tables S2–S4. Associations were directionally consistent across models, and multiple imputations yielded more precise estimates. Model discrimination was comparable (Harrell’s C = 0.82 for model 2; mean: 0.80 across imputed datasets for model 3).

Table 3.

Cox proportional hazards models to explore factors associated with mortality in the MEXHEMO cohort

Covariate / Category	Model 1: univariate (complete cases only)	Model 2: multivariate (missing categories)	Model 3: multivariate (Multiple imputation)
	HR (95% CI)	HR (95% CI)	HR (95% CI)
Vascular access
AVF	1 (ref.)	1 (ref.)	1 (ref.)
Tunneled catheter	1.46 (1.29–1.73)	1.23 (0.98–1.54)	1.28 (1.03–1.59)
Untunneled catheter	3.66 (2.95–4.56)	1.28 (0.99–1.67)	2.18 (1.70–2.80)
Femoral catheter	2.11 (1.46–3.07)	2.08 (1.41–3.07)	2.12 (1.44–3.12)
Other	1.67 (0.62–4.52)	1.27 (0.46–3.48)	1.99 (0.73–5.46)
Dialyzer reuse
No	1 (ref.)	1 (ref.)	1 (ref.)
Yes (without interaction term)	1.49 (1.29–1.73)	1.81 (1.40–2.34)	2.07 (1.64–2.61)
Yesa (with interaction term)	–	1.24 (0.86–1.79)	1.40 (0.98–1.99)
Reuse-center interactiona	–	1.04 (1.01–1.06)	1.04 (1.01–1.06)
Anemia
Hemoglobin < 9 g/dl	2.21 (1.71–2.87)	1.66 (1.26–2.19)	1.59 (1.21–2.07)
Hemoglobin 9 to < 10 g/dl	1.27 (0.95–1.72)	1.34 (0.98–1.83)	1.19 (0.89–1.61)
Hemoglobin 10 to < 11 g/dl	1.05 (0.77–1.45)	1.12 (0.81–1.55)	1.04 (0.76–1.41)
Hemoglobin 11 to < 12 g/dl	1 (ref.)	1 (ref.)	1 (ref.)
Social security coverage
Yes	0.60 (0.49–0.74)	0.65 (0.46–0.91)	0.50 (0.37–0.67)
No	1 (ref.)	1 (ref.)	1 (ref.)
Low flux dialysis
HF	1 (ref.)	1 (ref.)	1 (ref.)
LF	1.72 (1.42–2.09)	1.39 (0.99–1.96)	1.71 (1.28–2.28)

AVF, arteriovenous fistula; AVG, arteriovenous graft; CI, confidence interval; HR, hazard ratio; HF, high flux; LF, low flux.

Multivariate model includes all covariates from univariate analysis and is adjusted for age, sex, social security coverage, dialysis vintage, session frequency, ultrafiltration rate, diabetes, hypertension, and cardiovascular disease (not shown on table, Supplementary Tables S1, S2, and S3). AVG hazard ratio could not be estimated due to a limited number of cases.

^aDialyzer reuse HR in multivariate models shown after the inclusion of a dialyzer reuse-clinic site interaction term. P-values of < 0.05 were considered statistically significant.

Compared with AVF, all catheter types were associated with higher mortality. In the fully adjusted imputed model (model 3), femoral and nontunneled catheters showed the strongest associations (HR: 2.12, 95% CI: 1.44–3.12; and HR: 2.18, 95% CI: 1.70–2.80), whereas tunneled catheters remained significantly associated with mortality (HR: 1.28, 95% CI: 1.03–1.59).

Dialyzer reuse was consistently associated with higher mortality (HR: 1.49, 95% CI: 1.29–1.73; HR: 1.81, 95% CI: 1.40–2.34; and HR: 2.07, 95% CI: 1.64–2.61 in models 1, 2, and 3, respectively). However, this association was attenuated after adding a reuse-by-center interaction (model 2: HR: 1.24, 95% CI: 0.86–1.79; model 3: HR: 1.40, 95% CI: 0.98–1.99), with a significant interaction term (HR: 1.04, 95% CI: 1.01–1.06).

Social security coverage was associated with lower mortality in all models (HR: 0.60, 95% CI: 0.49–0.74; HR: 0.65, 95% CI: 0.46–0.91; and HR: 0.50, 95% CI: 0.37–0.67 in models 1, 2, and 3, respectively).

Low-flux dialysis was associated with increased risk in models 1 and 3 (HR: 1.72, 95% CI: 1.42–2.09; and HR: 1.71, 95% CI: 1.28–2.28). Severe anemia (Hb < 9 g/dl) was strongly associated with mortality, with attenuation after adjustment (HR: 2.21, 95% CI: 1.71–2.87 in model 1; HR: 1.66, 95% CI: 1.26–2.19 in model 2; HR: 1.59, 95% CI: 1.21–2.07 in model 3).

Albumin showed the largest effect sizes: compared with albumin ≥ 4.0 g/dl, HRs in model 3 were 3.12 for albumin < 3.2 g/dl, 2.03 for albumin 3.2 to <3.5 g/dl, and 1.40 for albumin 3.5 to < 4.0 g/dl (all P values < 0.05). Phosphorus was associated with mortality (HR: 1.33 for 5.5 to < 7.0 mg/dl and HR: 1.39 for ≥ 7.0 mg/dl in model 3; P < 0.05). Hyperkalemia was not found to be associated with mortality, in fact counterintuitively, serum potassium ≥ 6 mEq/l showed lower risk across models (HR: 0.59, 95% CI: 0.47–0.74; HR: 0.68, 95% CI: 0.53–0.86; HR: 0.69, 95% CI: 0.54–0.86 in models 1–3, respectively).

Older age was consistently associated with higher mortality (HR: 1.31 for 45–64 years, 1.63 for 65–74 years, and 1.66 for > 75 years; all P < 0.01). Male sex (HR: 1.27, P = 0.013) and cardiovascular disease (HR: 1.60, P < 0.001) were also independently associated with increased risk (model 3 estimates; full models in Supplementary Table 4).

Discussion

MEXHEMO is the largest cohort of chronic HD patients reported to date in Mexico. With 11,779 patients, it likely represents 10% to 15% of the national HD population. Most of our data (90%) come from patients with social security coverage receiving HD in outsourced private clinics, a setting estimated to account for approximately one-third of patients on maintenance HD nationwide.6, 7, 8, 9 Despite this partial representativeness, the study provides a valuable snapshot of clinical practices and outcomes and offers relevant insights to inform health care planning, policy development, and future research. In the following sections we provide our insights into the most relevant findings.

Mortality Rate and Standardized Comparisons

Considering reported deaths and patients lost to follow-up, we observed a mortality rate of 130 per 1000 patient-yrs. Because deaths could not be confirmed through death certificates or external registries, we consider this approach conservative and appropriate given the characteristics of our population. After age adjustment, mortality in the MEXHEMO cohort was significantly higher than that reported by the USRDS (SMR: 1.31, 95% CI: 1.25–1.38) and the Chilean HD Census (SMR: 1.57, 95% CI: 1.52–1.62).

The mean age of our population (52 ± 16 years) was notably lower than that of the reference populations, both of which were > 60 years.^15,16 Thus, a greater proportion of deaths occurred in younger age groups in MEXHEMO. This difference may partly reflect survivorship bias among incident HD patients in Mexico, where access to HD is limited and most patients covered by social security initiate renal replacement therapy with PD; as a result, a substantial proportion of incident HD patients may represent survivors of failed PD, although we lack individual-level data to confirm this. In addition, a disproportionately high burden of chronic kidney disease among younger individuals has been described in Mexico,^19,20 and the low annualized transplantation rate in our cohort (1.28%) may further contribute to differences in age distribution and survival patterns.

Given the high prevalence of adverse prognostic indicators in the MEXHEMO population—including catheter use,^21,22 short dialysis sessions,²³ anemia,^–24, 25, 26 and hypoalbuminemia²⁷— higher mortality, compared with countries with established registries where these indicators are actively targeted for improvement, is not unexpected. For instance, catheter use among prevalent patients is reported at 20% in the USA and 34% in Chile, with anemia prevalence (Hb < 10 g/dl) of 24% and 22%, respectively^14,16; whereas in our cohort, catheter use reached 70% and anemia was 50%. Moreover, we confirmed a strong association of these factors with mortality in the statistical analyses.

Although >90% of patients were insured through the national social security system, coverage was independently associated with lower mortality risk, consistent with previous reports¹⁰ and underscoring the importance of regular healthcare access for patients with ESKD in Mexico.

Although the SMR adjusts for age and USRDS comparisons used rates standardized by sex, cause of ESKD, and dialysis vintage, comparisons across health systems with different access to transplantation and supportive care remain inherently limited. In addition, ethnicity could not be accounted for in US comparisons, which is relevant given the lower mortality observed among Hispanic patients in the US.^28,29 The inclusion of the Chilean HD census as a Latin American benchmark helps contextualize our findings, and the higher SMR reported with this reference is consistent with an overall lower mortality expected in Hispanic populations.

Dialysis Prescription and Session Length

The mean HD session duration was 3.1 ± 0.27 hours, substantially shorter than that reported in the US (3.6 hours)³⁰ and Europe (3.9 hours),³¹ and an estimated average of 3.7 to 4.0 hours in Chile.³² Overcrowding in outsourced clinics, most operating 4 shifts and many 5 shifts/d, largely explains the widespread adoption of 3-hour sessions. Patients used to 3-hour sessions may be often resistant to increasing treatment duration, further reinforcing this practice. Although adequacy targets (spKt/V > 1.2) were reportedly achieved in 90% of patients, spKt/V measurements were not standardized, and treatment time is known to influence outcomes beyond solute clearance.^33,34

We did not find a consistent association between mortality and 3-hour HD sessions; however, the very small proportion of patients dialyzing for > 3.5 or 4 hours limited our ability to detect meaningful differences. Short session duration remains a relevant finding, given its potential effects on nutrition, anemia control, and volume management, and the growing evidence supporting longer HD treatments.^35,36

These findings highlight the need to reassess the incentives and structural drivers for short dialysis prescriptions, including reimbursement models. Public tenders, institutional requirements, and national regulations should be aligned to ensure sufficient operational capacity to deliver full 4-hour sessions. Without institutional and policy-level changes, sustained improvements in dialysis outcomes in Mexico will be difficult to achieve.

Dialyzer Reuse

Dialyzer reuse was reported in 43% of patients, reflecting a practice that persists in resource-limited settings despite being largely abandoned in many high-income countries. Although existing evidence—mostly from retrospective studies—suggests that reuse can be safe under strict standards,^–37, 38, 39 we observed higher mortality among patients undergoing reuse in all models. This association was attenuated after accounting for clinic-level variability, suggesting substantial heterogeneity in reuse practices.

This variability likely reflects the absence of strict regulations and standardized protocols, with practices ranging from automated, quality-controlled processes to manual reuse without adequate monitoring or traceability. Potential mechanisms linking reuse to adverse outcomes include membrane degradation, residual sterilant exposure, and increased risk of endotoxin translocation and bloodstream infections.^40,41 The lack of detailed information on reuse protocols, number of reuses, or related complications limits causal inference, but our findings support the need for updated national guidelines, quality control, and focused research on the safety of dialyzer reuse in Mexico.

Vascular Access

Vascular access remains a major challenge in Mexico. Our findings are consistent with prevous national reports.^–42, 43, 44 Catheter use was strikingly high despite a median HD vintage of 2 years and was consistently associated with increased mortality across all models. Notably, 16% of patients used nontunneled catheters, and 4% used femoral catheters, both associated with more than a 2-fold increase in mortality compared with AVFs. These findings are coherent with previous studies that have identified the use of catheters as an important risk factor for mortality.^19,20,45 Although catheter-related infections likely contribute to this risk, we were unable to evaluate this factor because of data limitations.

The overreliance on catheters reflects delayed referral, with most patients initiating HD through temporary access, as well as limited availability of vascular surgery services, evidenced by the minimal use of arteriovenous grafts. Structural factors play a role, particularly the reimbursement mechanisms for outsourced clinics in which vascular access creation and maintenance are bundled into the treatment payment, further reinforced by the fact that both tunneled catheters and AVFs are considered “permanent” accesses.

Promoting early vascular access planning, improving coordination between referring hospitals and dialysis providers, and prioritizing AVFs within quality metrics are essential steps to reduce long-term catheter dependence.

Laboratory Parameters

A large proportion of patients failed to meet the recommended laboratory targets for Hb, albumin, and phosphorus. Hypoalbuminemia (< 3.2 g/dl) was the strongest predictor of mortality, with adjusted HRs of 7.0 and 3.2 in multivariable models. Hyperphosphatemia and anemia were associated with increased risk. Particularly concerning was the finding that 30% of patients had Hb ≤ 9 g/dl, associated with an increase in mortality risk > 50%.

Although erythropoiesis-stimulating agents and iron therapy are covered by social security, medication availability is inconsistent and administration is often delayed due to logistical constraints, including the need for patients to go to the hospital to receive their doses rather than receiving them in the dialysis clinic. Iron deficiency likely contributes, because ferritin levels were < 500 ng/ml in 70% of patients and transferrin saturation < 30% in 65%, although insufficient data on iron supplementation and erythropoiesis-stimulating agent dosing precluded further analysis.

The inverse association between hyperkalemia and mortality likely reflects unmeasured confounding related to nutritional status and illustrates the limitations of single-point laboratory measurements.

Beyond their statistical associations, hypoalbuminemia and anemia represent actionable clinical targets, reflecting broader vulnerabilities such as malnutrition, inflammation, and suboptimal dialysis care. International initiatives, including the Dialysis Outcomes and Practice Patterns Study and the International Society of Renal Nutrition and Metabolism highlight the potential of systematic interventions—nutritional support and counseling; optimized erythropoiesis-stimulating agent use, and longer dialysis sessions—to improve outcomes, representing clear opportunities for quality improvement in Mexico.

Data Quality and System Gaps

Our findings reveal substantial gaps in data infrastructure. Data on albumin was missing in 16% of patients, and parathyroid hormone in 96%, pointing to deficiencies in routine monitoring. Improving data completeness is essential for quality assurance and effective risk stratification.

We did not report vascular access–related infections or associated mortality because of significant underreporting. Although exploratory analyses were conducted, data limitations prevented unbiased conclusions. Future studies should incorporate systematic infection surveillance and standardized cause-of-death reporting.

Limitations and Strengths

This study has several limitations. The cohort reflects patients with social security coverage and regular access to chronic HD and does not represent uninsured or rural populations, who face major barriers to renal replacement therapy and experience higher mortality, as previously reported by Valdez-Ortiz et al.¹⁰ Furthermore, our data came from only 7 States in the country, with 92% coming from Mexico City, State of Mexico, Querétaro, and Puebla. Voluntary clinic participation may have further introduced selection bias favoring better-resourced centers, limiting generalizability. Follow-up was limited to 6 months, restricting the assessment of long-term outcomes.

Mortality ascertainment relied exclusively on clinic records without linkage to civil registries, potentially leading to underreporting. To mitigate this, patients lost to follow-up were included in mortality estimates, which, though imperfect, likely approximates true mortality more closely. Cause-of-death data were unavailable, limiting cause-specific analyses.

We relied on single-point routine laboratory measurements and lacked data on several relevant variables, including socioeconomic status, ethnicity, nutritional indicators, body mass index, blood pressure, dialysate composition, and pharmacologic treatment, limiting adjustment for residual confounding.

Despite these limitations, the study’s strengths include its large sample size, multicenter design, and robust statistical approach. To our knowledge, this is the first large-scale analysis of dialyzer reuse in Mexico and provides essential epidemiological data in the absence of a national registry.

MEXHEMO represents the most comprehensive evaluation to date of chronic HD patients in Mexico. Mortality was higher than in the US and Chile after age adjustment, and the high prevalence of anemia, hypoalbuminemia, dialyzer reuse, and catheter use identifies clear targets for intervention. These findings support the need for national policies focused on standardizing care quality and addressing structural inequities in access and outcomes.

Appendix

List of the MEXHEMO Collaborative Group

Dr. Diana Maldonado Tapia, Centro Médico 20 de Noviembre ISSSTE.

Dr. Héctor Mayorga Madrigal, Clínica de Nefrología y Diálisis Querétaro.

Dr. María Teresa Gutiérrez González, Hospital General de Zona no. 30 IMSS.

Dr. Eduardo Guerrero Hinzpeter, Hospital General de México “Dr. Eduardo Liceaga”

Dr. Diana Carolina Sánchez Guerrero, Christus Muguerza Alta Especialidad.

Dr. Ramón Medina González, Hospital Civil de Guadalajara Fray Antonio Alcalde.

Dr. Edgar Marcelo Arellano Torres, Christus Muguerza, Hospital Sur.

Dr. Martha Yaneth Cantú Hinojosa, Hospital San Felipe.

Dr. Cristina Lourdes Lerma Narváez, Hospital Christus Muguerza del Parque.

Dr. Alejandro Valdés Cepeda, Tec-Salud.

Dr. Rafael Baizabal Olarte, Renaluz, Clínica de Nefrología sede Millennium.

Dr. Gilberto Galván Ramírez, Hospital Darío Fernández ISSSTE.

Dr. Francisco Hidalgo Alquicira, Hospital Regional Lic. Adolfo López Mateos ISSSTE.

Dr. Jorge Anselmo Peña Pérez, Hospital General Tláhuac SEDESA.

Dr. José Francisco Bueno Hernández, Centro de Asistencia Renal S.A. de C.V. “CARE”.

Dr. Abraham Santos Ontiveros, SERME.

Dr. Marco Carmona Escamilla, Hospital Central Petróleos Mexicanos PEMEX.

Dr. Ivanna Rocha Mercado, Unidad de Hemodiálisis HAVAN.

Dr. Patricia Coral Ruiz Palacios, Soporte Renal Integral.

Dr. Enzo Vásquez Jiménez, Hospital Juárez de México.

Dr. Adriana Capurro Ceballos, Hospital Ángeles del Carmen.

Disclosure

JMAI is employed by Médica Santa Carmen, a company that provides hemodialysis treatment services. JMRG, JA, and JCPR are employed by CEDIASA, a company also engaged in providing hemodialysis treatment. All the other authors declared no competing interests.

Data Availability Statement

The data that support the findings of this study contain confidential patient information and cannot be made publicly available to protect participant privacy in accordance with institutional and ethical regulations. However, deidentified data may be shared upon reasonable request from qualified researchers, subject to approval by the authors and the local ethics committee.

Supplementary Material

Supplementary File (PDF)

Acknowledgments. Table S1. MEXHEMO crude and mortality rates. Table S2. Cox model 1: univariate, complete cases only. Table S3. Cox model 2: multivariate with “missing categories.” Table S4. Complete Cox proportional hazards multivariate analysis with “missing category” to account for missing data (model 2). STROBE Checklist.