Acute kidney phenotypes in France: nationwide annual first-hospitalization rates and phenotype-specific epidemiology from administrative health data

Abstract

Background

Acute kidney phenotypes are a major cause of hospitalization, but nationwide data on their hospitalization rates and sociodemographic characteristics are limited. We conducted a nationwide hospitalization-based study of acute kidney phenotypes identified from administrative health data in France.

Methods

We performed a nationwide study using data from the French National Health Data System between 2019 and 2021. Hospitalizations with a principal discharge diagnosis corresponding to one of four acute kidney phenotypes - acute kidney injury (AKI), glomerulopathy, tubulointerstitial/acute pyelonephritis (TIN/AP), or nephrolithiasis - collectively referred to as acute kidney phenotypes (AKP) were identified in the French national hospital discharge database using International Classification of Diseases, 10th Revision (ICD-10) codes. Annual first-hospitalization rates were age and sex-standardized to the 2021 French population and expressed per 1,000,000 inhabitants with 95% confidence intervals (CIs).

Results

Among 624,130 patients hospitalized with an AKP, nephrolithiasis accounted for 49.8% of cases, TIN/AP for 34.2%, AKI for 12.3%, and glomerulopathy for 3.7%. Age- and sex-standardized annual first-hospitalization rates (per 1,000,000 inhabitants) by phenotype are reported for each calendar year. For glomerulopathy, rates were 132 (95% CI 130–135) in 2019, 115 (112–117) in 2020 and 116 (113–119) in 2021. AKI rates were 428 (423–433) in 2019, 409 (404–414) in 2020 and 428 (423–433) in 2021. Nephrolithiasis rates were 1,782 (1,772-1,792), 1,700 (1,690-1,710) and 1,737 (1,727-1,746), respectively. TIN/AP rates were 1,162 (1,154-1,170) in 2019, 1,063 (1,055 − 1,071) in 2020 and 1,039 (1,031 − 1,047) in 2021. TIN/AP showed a bimodal age distribution (20.3% <10 years; 27.3% ≥80 years) and a marked female predominance (73.5%), consistent with the predominance of acute pyelonephritis within this claims-based phenotype (84%).

Conclusion

This nationwide study provides the first hospitalization-based estimates of four acute kidney phenotypes in France and highlights distinct epidemiological profiles across subtypes. These findings support targeted prevention and monitoring of high-risk patients and provide a framework for future pharmacoepidemiological studies of drug-associated nephrotoxicity.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12882-026-05206-3.

Introduction

Acute kidney diseases represent a growing public health concern due to their increasing incidence, associated morbidity, and economic impact, particularly in elderly and hospitalized populations [1–4]. Acute kidney presentations encountered in routine clinical care and administrative healthcare databases are clinically heterogeneous and may involve glomerular, tubular, interstitial, hemodynamic, or obstructive mechanisms [5]. Existing nephrology classifications such as acute kidney injury (AKI) and acute kidney disease (AKD) remain clinically essential but were not specifically designed for operational phenotyping in large-scale pharmacoepidemiologic studies using hospitalization claims data. To facilitate pharmacoepidemiologic research using nationwide hospitalization claims databases, we developed an operational framework of kidney-related hospitalization phenotypes identifiable through ICD-10 coding. This framework was conceptually inspired by the phenotype-oriented approach proposed by Mehta et al. for drug-induced kidney disease [5], while adapting it to the constraints and objectives of administrative healthcare database research. Rather than proposing a new clinical classification of acute kidney diseases, our objective was to identify clinically meaningful and pharmacoepidemiologically actionable kidney phenotypes that could serve as potential safety outcomes in future nephrotoxicity studies. Accordingly, we examined four broad hospitalization phenotypes reflecting distinct patterns of kidney involvement: glomerulopathy, tubulointerstitial/infectious kidney disease, acute kidney injury, and obstructive/urolithiasis-related kidney events, collectively referred to as acute kidney phenotypes (AKP).

The exact frequency of acute kidney-related hospitalization phenotypes remains difficult to establish because of heterogeneous clinical presentations, variability in coding practices, coexistence of multiple kidney and systemic disorders, and, for some entities, the requirement for histological confirmation. Internationally, AKI incidence is 1–5/1,000 person-years, affects 1.9–7.2% of hospitalized patients and 20–50% in ICU [6–8]. Glomerulopathy incidence spans 0.2–2.5/100,000 [9]. TIN appears in 1–5% of general kidney histology and 10–35% of biopsies done for AKI [10]. Nephrolithiasis affects 10% of adults, with regional variability [11].

In France, data are fragmented: nephrolithiasis affects 10–12% of the population [10, 12], AKI 15% of inpatients and 30% in intensive care unit according to a monocentric study [13], with 1% to 6% requiring dialysis; primary glomerular diseases had an estimated prevalence of about 7 per 1,000 inhabitants in western France [14]; and immuno-allergic acute interstitial nephritis 3% of AKI cases [15]. The prevalence of risk factors (e.g. hypertension, diabetes mellitus, heart failure, or autoimmune diseases) in AKP populations is still poorly characterized at national scale, despite recognized associations and iatrogenic contributors [16–19].

Administrative healthcare databases are not designed to capture detailed nephropathological diagnoses, but they can support large-scale identification of clinically relevant kidney hospitalization patterns suitable for epidemiological safety surveillance.

The French National Health Data System (Système National des Données de Santé, SNDS) enables nationwide, hospitalization-based estimates and patient profiling from routinely collected data (near-complete coverage, standardized coding) [20]. Our objective was to estimate annual first-hospitalization rates for four operational kidney-related hospitalization phenotypes (glomerulopathy, tubulointerstitial/acute pyelonephritis (TIN/AP), AKI, and nephrolithiasis) in France (2019–2021) and to describe the distribution of sociodemographic characteristics, lifestyle factors, and comorbidities overall and by phenotype using SNDS [5, 20–23]. This is a pre-etiological step to inform future studies on drug-induced nephrotoxicity.

Methods

We conducted a retrospective, descriptive, population-based study using routinely collected medico-administrative data from the SNDS over a three-year period from January 1, 2019, to December 31, 2021 [20]. The index date was defined as the date of the first PMSI-MCO hospitalization during the calendar year in which an AKP was recorded as the principal discharge diagnosis (DP/PD). Unless otherwise specified, hospitalizations refer to PMSI-MCO (Hospitalizations in medicine, surgery, and obstetrics wards), including day-case (0-night) stays. Emergency department visits not resulting in an MCO admission and outpatient-managed events are not captured in PMSI-MCO.

Restricting case identification to the principal diagnosis was intended to maximize specificity and to focus on hospitalizations primarily driven by an acute kidney phenotype, yielding conservative annual first-hospitalization estimates; however, this approach may under-ascertain AKPs, particularly AKI, when coded as secondary diagnoses.

Data sources, data cleaning, and missing data

The SNDS comprises (i) DCIR (Datamart de Consommation Inter-Régime), corresponding to claims for ambulatory care and dispensations; (ii) PMSI (Programme de Médicalisation des Systèmes d’Information), corresponding to hospital discharge abstracts with ICD-10 diagnoses and procedures across MCO (medicine, surgery, obstetrics and dentistry), SSR (soins de suite et réadaptation, rehabilitation care), Psy (psychiatric hospitalizations), and HAD (Hospitalization à domicile, home hospitalization); and (iii) BCMD (Base de Causes Médicales de Décès), for cause of death.

We deduplicated records to one row per patient per year and applied consistency checks (valid age ranges by phenotype, non-negative LOS). Missingness was not imputed; variables with incomplete capture (e.g., the French Deprivation Index, FDep) included an “Unknown” category.

Study population: Eligibility, case definitions, and exclusions

We identified all hospitalized patients with a PD corresponding to one of four kidney phenotypes: glomerulopathy (N00-N06, N08), TIN/AP (N10, N11-N12, N14-N16), AKI (N17, N19), or nephrolithiasis (N20-N23) (supplementary Appendix S2). All ages were eligible. Patients whose only diagnosis was a hereditary nephropathy (N07) or an obstructive and reflux uropathy (N13) were excluded (Fig. 1).

Fig. 1.

Patient selection and phenotype assignment (AKP, PMSI-MCO, 2019–2021). SNDS: French National Health Data System. PMSI-MCO: hospital discharge database for medicine, surgery and obstetrics. AKP: acute kidney phenotype. PD: Principal discharge diagnosis. AKP principal diagnosis code sets: glomerulopathy (N00-N06/N08), tubulointerstitial nephritis / acute pyelonephritis (N10, N11-N12, N14 N16), acute kidney injury (N17, N19), nephrolithiasis (N20-N23). Exclusion diagnoses: hereditary nephropathy (N07) and obstructive/reflux uropathy (N13). Expert-informed classification algorithm leads to final kidney phenotype assignment: - If kidney biopsy: phenotype anchored by the biopsy stay; - Otherwise: priority rules to resolve discordant PDs (Supplementary Appendix S1)

Although Mehta et al.’s tubular phenotype primarily captures tubular dysfunction that may occur without overt AKI, our claims-based TIN/AP phenotype was defined to capture diagnoses explicitly consistent with tubulointerstitial inflammation when recorded as discharge diagnoses, most notably acute TIN/acute pyelonephritis (N10) and related tubulointerstitial entities (N11-N16) (Supplementary Appendices S2). In contrast, acute tubular necrosis and functional forms of AKI were captured under the AKI phenotype through ICD-10 AKI codes and the prioritization algorithm (e.g., N17.0 and N17.9 when applicable; Supplementary Appendix S1). Because serum creatinine and urine output are not available in SNDS, AKI severity could not be staged using KDIGO criteria; ICD-10 coded AKI may therefore correspond to any KDIGO stage, with principal-diagnosis hospitalizations likely corresponding to clinically more severe cases.

Because patients can experience multiple AKP-coded hospitalizations, we implemented an expert-informed classification algorithm (Supplementary Appendix S1) to assign one final kidney phenotype per patient (Supplementary Table S5). Briefly, in patients with multiple AKP-coded stays, kidney biopsy (when present) anchored the phenotype; otherwise, pre-specified priority rules resolved discordance (N10 > N00-N06/N08 > N11-N12/N14-N16 > N17.0 > N17.9 > N17.1-N17.8/N19 > N20-N23). As a sensitivity analysis, an alternative prioritization hierarchy was also applied: N00-N06/N08 > N17.0 > N10 > N11-N12/N14-N16 > N17.1-N17.8/N17.9/N19 > N20-N23.

An exploratory sensitivity analysis was additionally performed to assess the influence of dominant phenotypes on the overall distribution of operational kidney hospitalization phenotypes.

For single-stay patients in 2019–2021, we screened ± 2 years around the index date to check for additional AKP-coded hospitalizations relevant to final phenotype assignment (Supplementary Appendix S1 and Supplementary Table S1-S2).

Variables

We extracted age, sex, region of residence, and length of hospital stay (LOS). For patients with multiple AKP hospitalizations in a year, LOS was summarized at patient level (median [IQR]). Social deprivation was measured using the French Deprivation Index (FDep 2015) in quintiles (1 = least deprived, 5 = most deprived); an “Unknown” category was retained when linkage was not possible. Lifestyle and comorbidities were identified using the G11 algorithms from the French national disease-mapping methodology (2015–2022, all schemes) [24]. We considered tobacco use, alcohol use, diabetes, antihypertensive treatment, heart failure, coronary artery disease, stroke, cancers (all types), rheumatoid arthritis, systemic lupus erythematosus, liver disease, pancreatic disease, pulmonary embolism, chronic inflammatory bowel disease, chronic respiratory disease, hereditary metabolic disorders or amyloidosis, peripheral arterial disease and HIV infection. Comorbidity and lifestyle flags corresponded to the calendar year of the index AKP, as defined in G11.

Bias

Misclassification and under-ascertainment are possible due to reliance on ICD-10 codes and restriction to principal diagnosis. This choice improves specificity but may differentially under-capture AKPs that are frequently recorded as secondary diagnoses, particularly AKI (often secondary to sepsis, acute heart failure, shock or other acute illnesses). Consequently, our phenotype distribution should not be interpreted as the relative burden of all AKI versus nephrolithiasis or TIN/AP in hospital settings; rather, it reflects hospitalizations where the AKP is coded as the main reason for care. Besides, coding practices may differ across hospitals, and laboratory and biopsy data are not available in the SNDS. We mitigated discordance by applying an expert-based prioritization algorithm and screening ± 2 years around the index date. Finally, the COVID-19 pandemic likely affected care-seeking and coding in 2020–2021, which may have influenced observed hospitalization patterns.

Study size

No formal sample size calculation was performed, as there was no hypothesis testing and no sampling. All eligible SNDS hospitalizations with an AKP as principal diagnosis in 2019–2021 were included. Study size was therefore determined by national coverage and data availability.

Statistical analysis

We performed descriptive analysis: means (SD), medians (IQR), and annual first-hospitalization rates (95% CI) for continuous variables; and counts and proportions (n, %) for categorical variables. For each calendar year, annual first-hospitalization rate estimates were based on the first PMSI-MCO hospitalization in which an acute kidney phenotype was recorded as the principal discharge diagnosis, after excluding patients with any acute kidney-coded PMSI-MCO hospitalization, irrespective of phenotype, during the preceding two-year look-back period. These estimates therefore represent annual first-hospitalization rates, not true disease incidence or first-ever occurrence of the underlying kidney condition. For each year, annual first-hospitalization rates were age- (5-year bands) and sex-standardized to the 2021 French population as the reference. Data on the French population by age and sex for each year were retrieved from the INSEE (National Institute of Statistics and Economic Studies) website [25]. Rates were expressed per 1,000,000 inhabitants with 95% confidence intervals computed under a Poisson assumption. All statistical analyses were performed using SAS Enterprise Guide, version 8.

Ethics and regulatory approval

This study was conducted using SNDS data, a national medico-administrative database containing pseudonymized individual-level health information. EPIPHARE’s access to the SNDS was granted by French regulatory authorities in accordance with Decree No. 2016 − 1871 (December 26, 2016), and Articles R.1461-13 and R.1461-14 of the French Public Health Code [26].

The study was registered on May 3, 2024, in the EPIPHARE (ANSM-Cnam) study registry, in compliance with SNDS research requirements (Study Declaration Reference: T-2024-05-568).

Results

Between 2019 and 2021, a total of 624,130 patients were hospitalized in France with a PD of an acute kidney phenotype (AKP). The most frequent phenotype was nephrolithiasis (310,725; 49.8%), followed by tubulointerstitial nephritis and acute pyelonephritis (TIN/AP; 213,685; 34.2%), acute kidney injury (AKI; 76,746; 12.3%) and glomerulopathy (22,974; 3.7%) (Table 1). Patient selection and phenotype assignment using the expert algorithm are detailed in Appendices 1–3 and summarized in a RECORD-compliant figure (Fig. 1). Among multi-stay patients, PD concordance was high for nephrolithiasis and TIN/AP and lower for AKI and glomerulopathy; details of the classification algorithm and concordance metrics are provided in Supplementary Appendix S1 and Supplementary Tables S1-S2, together with the sensitivity analysis of the phenotype assignment algorithm (Supplementary Table S3). An exploratory sensitivity analysis assessing the influence of dominant phenotypes on the overall distribution of operational kidney hospitalization phenotypes is presented in Supplementary Table S4. Annual first-hospitalization rates for each acute kidney phenotype were estimated for each calendar year. For glomerulopathy, the age- and sex-standardized annual first-hospitalization rates (per 1,000,000 inhabitants), was 132 [95% CI: 130–135] in 2019, 115 [112–117] in 2020, and 116 [113–119] in 2021. For acute kidney injury (AKI), age- and sex-standardized annual first-hospitalization rates were 428 [423–433] in 2019, 409 [404–414] in 2020, and 428 [423–433] in 2021. For nephrolithiasis, the respective rates were 1,782 [1,772-1,792], 1,700 [1,690-1,710], and 1,737 [1,727-1,746]. For TIN/AP, the annual first-hospitalization rates were 1,162 [1,154-1,170] in 2019, 1,063 [1,055 − 1,071] in 2020, and 1,039 [1,031 − 1,047] in 2021 (Fig. 2).

Table 1.

Characteristics of the study population

	Nephrolithiasis		Tubulointerstitial Nephritis / acute pyelonephritis		Acute Kidney Injury		Glomerulopathy		Total
	n	%	n	%	n	%	n	%	n	%
Patients	310,725	49.8%	213,685	34.2%	76,746	12.3%	22,974	3.7%	624,130	100%
Single-stay patients	172,904	55.6%	165,835	77.6%	62,461	81.4%	16,427	71.5%	417,627	66.9%
Two hospital stays	85,921	27.6%	35,663	16.7%	10,931	14.2%	4794	20.9%	137,309	22.0%
Three hospital stays	51,900	16.7%	12,187	5.7%	3354	4.4%	1753	7.6%	69,194	11.1%
Median Length of Stay (days)	1 (0–1)		3 (2–6)		6 (2–12)		3 (2–7)		2 (0–4)
Age (year) - mean (std)	53.4 (17)		51.3 (32)		72.5 (16)		56.5 (25)		55.2 (24)
Age (year) - median (Q1 - Q3)	54 (41–66)		61 (21–81)		75 (65–84)		63 (39–76)		59 (39–74)
Age
< 10	584	0.2%	43,327	20.3%	256	0.3%	1522	6.6%	45,689	7.3%
10–19	5136	1.7%	8371	3.9%	589	0.8%	1165	5.1%	15,261	2.4%
20–29	23,686	7.6%	14,058	6.6%	1048	1.4%	1306	5.7%	40,098	6.4%
30–39	40,661	13.1%	12,328	5.8%	1718	2.2%	1820	7.9%	56,527	9.1%
40–49	53,851	17.3%	11,936	5.6%	2736	3.6%	1920	8.4%	70,443	11.3%
50–59	67,113	21.6%	15,022	7.0%	6275	8.2%	2767	12.0%	91,177	14.6%
60–69	62,496	20.1%	21,337	10.0%	14,411	18.8%	3864	16.8%	102,108	16.4%
70–79	40,295	13.0%	29,030	13.6%	20,774	27.1%	4295	18.7%	94,394	15.1%
≥ 80	16,903	5.4%	58,276	27.3%	28,939	37.7%	4315	18.8%	108,433	17.4%
Sex (Female)	108,162	34.8%	157,015	73.5%	29,651	38.6%	7465	32.5%	302,293	48.4%
Regions
Ile de France	38,304	12.3%	33,543	15.7%	12,693	16.5%	4463	19.4%	89,003	14.3%
Auvergne-Rhône-Alpes	39,316	12.7%	25,503	11.9%	9921	12.9%	2987	13.0%	77,727	12.5%
Hauts-de-France	28,025	9.0%	23,203	10.9%	9266	12.1%	2083	9.1%	62,577	10.0%
Nouvelle-Aquitaine	33,937	10.9%	20,276	9.5%	5751	7.5%	1963	8.5%	61,927	9.9%
Provence-Alpes-Côte d’Azur	31,559	10.2%	18,723	8.8%	5914	7.7%	1600	7.0%	57,796	9.3%
Occitanie	31,932	10.3%	16,466	7.7%	6555	8.5%	1923	8.4%	56,876	9.1%
Grand Est	26,711	8.6%	17,291	8.1%	7548	9.8%	1904	8.3%	53,454	8.6%
Normandie	14,657	4.7%	13,678	6.4%	3208	4.2%	1040	4.5%	32,583	5.2%
Pays de la Loire	16,768	5.4%	11,403	5.3%	3106	4.0%	1151	5.0%	32,428	5.2%
Bourgogne-Franche-Comté	14,534	4.7%	8885	4.2%	3145	4.1%	969	4.2%	27,533	4.4%
Bretagne	11,376	3.7%	10,666	5.0%	2965	3.9%	967	4.2%	25,974	4.2%
Centre Val-de Loire	12,178	3.9%	7185	3.4%	3942	5.1%	845	3.7%	24,150	3.9%
Corse	2946	0.9%	985	0.5%	370	0.5%	70	0.3%	4371	0.7%
DROM	8482	2.7%	5878	2.8%	2362	3.1%	1009	4.4%	17,731	2.8%
Social deprivation index (quintiles)
1 (the least deprivation)	52,534	17.2%	35,975	17.0%	13,158	17.4%	4200	18.6%	105,867	17.3%
2	60,103	19.7%	38,261	18.2%	12,925	17.1%	4095	18.1%	115,384	18.8%
3	61,613	20.2%	41,532	19.7%	14,545	19.3%	4196	18.6%	121,886	19.9%
4	61,387	20.1%	42,253	20.7%	15,278	20.3%	4359	19.3%	123,277	20.1%
5 (the most deprivation)	60,939	20.0%	46,687	22.2%	17,182	22.8%	4731	20.9%	129,539	21.1%
Unknown	8398	2.7%	5806	2.8%	2340	3.1%	1000	4.4%	17,544	2.9%
Lifestyle-related variables
Alcohol abuse	5714	1.8%	7122	3.3%	8189	10.7%	1231	5.4%	22,256	3.6%
Tobacco use	25,464	8.2%	17,093	8.0%	13,048	17.0%	2971	12.9%	58,576	9.4%

Fig. 2.

Age- and sex-standardized annual first-hospitalization rates of acute kidney phenotypes, France 2019–2021 (per 1,000,000 inhabitants). INSEE population by age and sex, year 2019 (66,988,403 inhabitants): https://www.insee.fr/fr/statistiques/6455187?geo=FE-1&sommaire=6455209. INSEE population by age and sex, year 2020 (67,162,154 inhabitants): https://www.insee.fr/fr/statistiques/7633062?geo=FE-1&sommaire=7633086. INSEE population by age and sex, year 2021 (67,408,052 inhabitants): https://www.insee.fr/fr/statistiques/8200787?geo=FE-1&sommaire=8200811

Between 5% and 12% of patients hospitalized for an AKP in a given year had been hospitalized for an AKP in the previous year, and 3% to 7% had been hospitalized two years earlier (Table 2).

Table 2.

Annual first AKP-coded PMSI-MCO hospitalizations after a two-year AKP-free look-back period, 2019–2021

	Glomerulopathy		Acute Kidney Injury		Nephrolithiasis		Tubulointerstitial Nephritis/ Acute pyelonephritis		Total
	n	%	n	%	n	%	n	%	n	%
1 First hospitalizations included in annual rate calculations (2019)	8744	3.8%	28,013	12.0%	118,572	50.9%	77,736	33.4%	233,065	100%
2 Patients excluded because of hospitalization during year n − 1	672	7.7%	1698	6.1%	15,285	12.9%	4901	6.3%	22,556	9.7%
3 Patients excluded because of hospitalization during year n − 2	396	4.5%	926	3.3%	8323	7.0%	2994	3.9%	12,639	5.4%
First hospitalizations included in annual rate calculations (2020)	7638	3.5%	27,160	12.4%	113,734	51.7%	71,332	32.4%	219,864	100%
Patients excluded because of hospitalization during year n − 1	648	8.5%	1682	6.2%	14,470	12.7%	4708	6.6%	21,508	9.8%
Patients excluded because of hospitalization during year n − 2	354	4.6%	931	3.4%	7894	6.9%	2912	4.1%	12,091	5.5%
First hospitalizations included in annual rate calculations (2021)	7815	3.5%	28,831	12.9%	117,054	52.3%	70,041	31.3%	223,741	100%
Patients excluded because of hospitalization during year n − 1	643	8.2%	1654	5.7%	14,991	12.8%	4369	6.2%	21,657	9.7%
Patients excluded because of hospitalization during year n − 2	376	4.8%	1001	3.5%	8548	7.3%	2966	4.2%	12,891	5.8%

¹ Number of patients included in the calculation of annual first-hospitalization rates, defined as patients with a first hospitalization for an acute kidney phenotype (AKP) during the calendar year under consideration and no previous hospitalization for any AKP during the preceding two-year look-back period

² Patients excluded from annual first-hospitalization rate calculations because of a hospitalization for an acute kidney phenotype during the preceding calendar year (year n − 1)

³ Patients excluded from annual first-hospitalization rate calculations because of a hospitalization for an acute kidney phenotype during the second preceding calendar year (year n − 2)

Rows corresponding to years n-1 and n-2 describe AKP-coded hospitalizations identified during the two-year look-back period and excluded from the numerator of annual first-hospitalization rates. These rows are provided to document the impact of the look-back period and should not be interpreted as part of the annual rate numerator. Patients with AKP-coded hospitalizations in both look-back years may contribute to both rows

For TIN/AP, a bimodal age distribution was observed, with a median of 61 years [21–81]; 20.3% of patients were under 10 years of age, and 27.3% were aged 80 years or older. This pattern was largely driven by acute pyelonephritis, which accounted for the vast majority of TIN/AP diagnoses (84% coded as N10), explaining both the pediatric peak and the overrepresentation of very old patients. Other tubulointerstitial diagnoses were rare and contributed marginally to the overall age distribution. Among patients with glomerulopathy, 57.5% were between 50 and 79 years old. The mean age was 56.6 years (SD: 25), and the median was 63 years [39–76]. In the case of nephrolithiasis, 59% of patients were aged 40–69 years. The mean age was 53.4 years (SD: 17), and the median was 54 years [41–66] (Table 1; Fig. 3).

Fig. 3.

Distribution of acute kidney phenotypes by age group at index hospitalization. Bars show the proportion of each acute kidney phenotype by age category. The bimodal age distribution observed for tubulointerstitial nephritis and acute pyelonephritis (TIN/AP) is largely driven by acute pyelonephritis (N10), which accounts for the majority of TIN diagnoses, particularly in pediatric and very old populations

Sex ratios were also phenotype-specific. Women accounted for 73.5% of TIN/AP cases (157,015 out of 213,685), 38.6% of AKI cases (29,651 out of 76,746), 32.5% of glomerulopathy cases (7,465 out of 22,974), and 34.8% of nephrolithiasis cases (108,162 out of 310,725).

The median hospital stay was 6 days [2–12] for AKI, 3 days [2–7] for glomerulopathy, 3 days [2–6] for TIN/AP, and 1 day [0–1] for nephrolithiasis.

Comorbidity profiles differed markedly across AKP (Fig. 4). Overall, antihypertensive treatment was present in 40.3% of patients, followed by diabetes (17.6%), cancer (14.8%), and chronic respiratory disease (10.3%). Coronary disease and heart failure were reported in 9.7% and 8.0% of patients, respectively. Among autoimmune diseases, systemic lupus erythematosus was found in 1.4% and rheumatoid arthritis in 1.1% of cases. Phenotype-specific distributions are shown in Fig. 4 (and detailed in Supplementary Table S6).

Fig. 4.

Comorbidities were identified using the French disease-mapping G11 algorithms and correspond to the calendar year of the index AKP (see Methods). Abbreviations: AKP, acute kidney phenotype; SNDS, French National Health Data System

Discussion

The objective of this study was to identify acute kidney phenotypes (AKPs) and to describe their epidemiology using comprehensive data from the French National Health Data System (SNDS). During the study period, 624,130 patients hospitalized for an AKP were identified. Nearly half of these patients presented with nephrolithiasis (49.8%), one-third with tubulointerstitial nephritis and acute pyelonephritis (TIN/AP, 34.2%), while acute kidney injury (AKI) and glomerulopathy accounted for 12.3% and 3.7% of cases, respectively.

Comparison across the four AKP subtypes revealed several specific features. The highest median age was observed in patients with AKI (75 years [65–84]), who also had the longest median length of hospital stay (6 days [2–12]). Women were markedly overrepresented in TIN/AP (73.5%) and less so in the other AKP categories. Finally, AKI was associated with the greatest burden of comorbidities, reflecting the complex clinical profile of this subgroup.

Comparison with the literature

Our findings, consistent with the literature [27, 28], suggest that the advanced age of patients with AKI may reflect population aging, physiological frailty, polypharmacy, and the increased prevalence of comorbid conditions [29–33].

The bimodal age distribution observed for TIN/AP (pediatric and very elderly populations) is consistent with the heterogeneity of its etiologies, which include infections, medications, and systemic diseases [34–37]. Similarly, the female predominance observed in TIN/AP (73.5%) most likely reflects the fact that, in our claims-based definition, the TIN/AP phenotype was largely driven by acute pyelonephritis (N10 being the predominant code). Acute pyelonephritis and UTIs are well known to be more frequent in women, and hospitalization generally concerns more severe presentations or patients with risk factors (e.g., older age, comorbidities, pregnancy) [38, 39]. In contrast, non-infectious acute interstitial nephritis and other tubulointerstitial entities were much less frequent in our data and therefore likely contributed less to the marked female predominance observed in this phenotype.

This predominance of infectious presentations also likely influenced the overall distribution of hospitalization-based kidney phenotypes observed in the study. Although AKI accounted for a lower proportion of overall hospitalization-based kidney phenotypes compared with nephrolithiasis and tubulointerstitial/acute pyelonephritis presentations, exploratory analyses showed that AKI became the predominant phenotype once highly prevalent obstructive and infectious phenotypes were sequentially excluded. This suggests that the apparent relative underrepresentation of AKI in the overall distribution was largely driven by the predominance of nephrolithiasis- and acute pyelonephritis-related hospitalizations within the operational framework.

Antihypertensive treatment was the most frequent therapy in the cohort (40.3%), with diabetes (17.6%) and cancer (14.8%) being the most common comorbidities. These conditions are frequently associated with impaired kidney function and are commonly reported in populations at risk for acute kidney-related hospitalizations, either due to underlying disease mechanisms or to treatment-related nephrotoxicity [32, 40–42]. Cardiovascular diseases (coronary artery disease: 9.7%; heart failure: 8.0%) and chronic respiratory diseases (10.3%) were also highly prevalent, consistent with the high burden of comorbid conditions observed in hospitalized patients with acute kidney phenotypes [43–46]. Finally, the prevalence of autoimmune diseases (systemic lupus erythematosus: 1.4%; rheumatoid arthritis: 1.1%) was low in our cohort, but their association with glomerulopathy and tubulointerstitial kidney disorders is well documented [18, 47–49].

Lessons learned and public health implications

Several important observations emerge from this work. The heterogeneity of epidemiological profiles across AKP subtypes highlights the importance of considering phenotype-specific approaches in future prevention, screening, and management strategies. Patients hospitalized with AKI were characterized by advanced age and a high comorbidity burden, suggesting a particularly vulnerable subgroup frequently encountered in internal medicine, geriatrics, and intensive care settings. The marked female predominance observed in TIN/AP in our claims-based definition may be consistent with a contribution of severe urinary tract and pyelonephritis-related hospitalizations within this phenotype.

The high prevalence of antihypertensive treatment, diabetes, cancer, and cardiovascular disease among patients with AKP is consistent with the important comorbidity burden observed in these populations. These findings may support closer coordination between nephrologists and other specialists to optimize follow-up and reduce kidney-related risk in these populations.

From a public health perspective, hospitalization-based acute kidney phenotypes may serve as markers of medical vulnerability and health-care utilization. Prevention strategies could include improved management of underlying chronic diseases, earlier identification of patients at risk, and reinforced surveillance during hospitalization.

Research perspectives

Our findings open several avenues for further investigation:

• Characterizing post-AKP trajectories, particularly the risk of progression to chronic kidney disease.

• Performing standardized mortality analyses and multivariable prognostic modeling to better characterize outcome heterogeneity and risk profiles across acute kidney phenotypes.

• Evaluating the economic burden of AKP management to inform healthcare funding and organizational decisions;

• Exploring drug-induced nephrotoxicity by analyzing each AKP subtype as a distinct endpoint rather than as part of a composite outcome, so as not to mask effects in less frequent entities.

This latter point is illustrated by the analysis of sex distribution: while women accounted for 48.4% of AKP overall, this aggregate figure conceals substantial disparities across subtypes. Women constituted 73.5% of TIN/AP cases but only about one-third of the other entities. A more granular analysis of AKP subtypes is therefore essential to refine our understanding of their determinants and to improve prevention and management strategies. Moreover, a two-year look-back period substantially reduces prevalent case carryover, thereby improving the specificity of analyses focused on incident AKP determinants.

Contribution and strengths of the study

This study offers several original contributions compared to existing literature. First, it represents one of the first large-scale analyses using data from the French National Health Data System (SNDS) to examine four acute kidney phenotypes (AKPs) in the general population. With a sample of over 600,000 patients, it provides the first population-level estimates of their frequencies.

Second, the multidimensional approach, which integrates sociodemographic, behavioral and clinical comorbidity data enables a comprehensive characterization of patients. This holistic analysis contributes to a better understanding of the associated factors and the heterogeneity of acute kidney pathologies. To our knowledge, few national studies have simultaneously examined the burden of AKP across the general population using administrative data.

Limitations of the study

This study is based on data from the SNDS, a medical-administrative database. As such, results should be interpreted in light of potential under-ascertainment from coding-only case capture, inter-hospital coding variability, and pandemic-related disruptions; the absence of laboratory and biopsy validation is an inherent limitation of administrative data. The analysis only included hospitalizations with AKPs coded as the main diagnosis, which may underestimate the true burden of these conditions. This under-ascertainment is likely stronger for AKI than for nephrolithiasis or TIN/AP, because AKI is frequently coded as a secondary diagnosis to another acute condition; we nevertheless chose this conservative, specificity-oriented approach to preferentially capture hospitalizations primarily driven by acute kidney phenotypes; therefore, comparisons of phenotype proportions should be interpreted cautiously. Because the 2019–2021 study period includes two years substantially affected by the COVID-19 pandemic and only one pre-pandemic year, the annual first-hospitalization rates reported here should be interpreted as period-specific descriptive estimates rather than as evidence of secular trends. This study was not designed to compare rates across calendar years or to assess the impact of COVID-19 on acute kidney phenotype hospitalizations. Longer pre- and post-pandemic time series are required to characterize temporal patterns and to distinguish underlying epidemiological changes from pandemic-related disruptions in hospital admissions, case severity at admission, and outcomes. Finally, this work does not yet incorporate detailed data on drug exposure, which will be explored in a future analysis focusing on pharmacoepidemiological aspects and nephrotoxicity. These findings are generalizable to hospitalized populations in France captured in PMSI-MCO. They are not directly generalizable to acute kidney phenotypes managed exclusively in outpatient settings or coded as secondary diagnoses. International extrapolation should be cautious given differences in admission thresholds, coding practices (e.g., principal vs. secondary diagnosis), and health-care utilisation. The nationwide coverage of the SNDS and the consistent operational approach nevertheless support reproducible phenotype identification and provide a basis for future longer-term monitoring.

Conclusion

This epidemiological study, based on data from the French National Health Data System (SNDS), allowed for the characterization of patients with acute kidney phenotypes (AKI, glomerulopathy, tubulointerstitial nephritis / acute pyelonephritis, and nephrolithiasis), highlighting significant disparities by age, sex, region of residence, and comorbidities.

This study identifies operational kidney phenotypes that may serve as outcome measures for future nephrotoxicity studies and characterizes major confounding factors relevant for pharmacoepidemiologic analyses.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

AKP

Acute kidney phenotype

AKI

Acute kidney injury

BCMD

National database on medical causes of death (Base de Causes Médicales de Décès)

DCIR

Inter-Scheme healthcare consumption database (Datamart de Consommation Inter-Régime)

HAD

Hospitalization à domicile, home hospitalization

KDIGO

Kidney disease improving global outcomes

PD

Principal diagnosis

PMSI-MCO

Hospitalizations in medicine, surgery, and obstetrics wards

SNDS

French national health data system (Système National des Données de Santé, SNDS)

SSR

Soins de suite et réadaptation, rehabilitation care

TIN/AP

Tubulointerstitial nephritis and acute pyelonephritis

Author contributions

Mwamba and Grimaldi conceived and designed the study. Mwamba performed the data acquisition and statistical analyses. Gerard, Lombardi and Jourde-Chiche provided clinical expertise. Mwamba, Gerard, Lombardi, Zureik, Jourde-Chiche and Grimaldi contributed to the interpretation of the data. Mwamba and Grimaldi drafted the manuscript. Gerard, Lombardi, Zureik, Jourde-Chiche and Grimaldi critically revised the manuscript for important intellectual content. Zureik, Jourde-Chiche and Grimaldi supervised the study. All authors approved the final version of the manuscript.

Funding

This research received no specific external funding.”

Data availability

The datasets generated and/or analysed during the current study are not publicly available.Data sharing is not applicable. According to data protection and the French regulation, the authors cannot publicly release the data from the French national health data system (SNDS). However, any person or structure, publicor private, for- profit or nonprofit, is able to access SNDS data upon authorization from the French Data Protection Office (CNIL Commission Nationale de l’Informatique et des Libertés) to carry out a study, research, or an evaluation of public interest [https://www.snds.gouv.fr/SNDS/Processus-d-acces-aux-donnees].

Declarations

Ethics approval and consent to participate

Not applicable. This study was conducted using pseudonymized data from the French National Health Data System (SNDS). Access to the SNDS is regulated under French law and authorized for EPIPHARE under Decree No. 2016 − 1871 and Articles R.1461-13 and R.1461-14 of the Public Health Code. In accordance with French regulations, individual informed consent was not required because the study used routinely collected administrative data that were fully pseudonymized.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.