CBC-derived inflammatory indices predict adverse outcomes in geriatric emergency patients with acute infectious...

Acute infectious diarrhea remains a leading cause of morbidity and mortality worldwide. Similar to pediatric populations, geriatric patients represent a highly vulnerable group with increased hospitalization rates, prolonged length of stay, heightened dehydration susceptibility, diminished physiological reserve, and elevated mortality risk. While pediatric populations benefit from well-established clinical severity scores such as Vesikari and Clark scales, analogous prognostic tools specifically validated for geriatric patients with acute infectious diarrhea remain lacking. Specific inflammatory markers such as procalcitonin and C-reactive protein require additional testing and increase costs. Complete blood count (CBC)-derived inflammatory indices offer a practical alternative as they are inexpensive, universally available, and calculated from routine tests. This retrospective study utilized the MIMIC-IV database (2008-2019) to assess diagnostic accuracy of nine CBC-derived inflammatory indices (AISI, NPAR, HALP, NLR, SII, PLR, MLR, SIRI, PNI) for predicting adverse outcomes in geriatric patients (≥ 65 years) with acute infectious diarrhea. Patients were identified through two independent database queries using ICD-9 codes (008-009 infectious diarrhea, 003-005 bacterial/foodborne; n = 2,208 raw records) and ICD-10 codes (A02-A05 bacterial/foodborne, A08 viral enteritis, A09 infectious gastroenteritis unspecified; n = 1,868 raw records), which yielded overlapping results due to MIMIC-IV spanning both coding eras (2008-2012 ICD-9; 2012-2019 ICD-10); after cross-encounter deduplication retaining one index admission per patient, the non-redundant cohort comprised 3,934 patients. Exclusions (n = 1,087) comprised non-infectious gastroenteritis, inflammatory bowel disease, radiation enteritis, ischemic colitis, drug-induced diarrhea, immunosuppressed patients, missing laboratory data, and C. difficile recurrence. Among 2,847 immunocompetent geriatric patients, comorbidities included hypertension (32.8%), diabetes mellitus (18.6%), and chronic kidney disease (10.4%). Bacterial pathogens were identified in 276 (9.7%) and viral in 151 (5.3%; norovirus 64.1%, rotavirus 23.4%, adenovirus 12.5%). Hospital admission occurred in 668 (23.5%), ICU admission in 34 (1.2%), in-hospital mortality in 14 (0.5%), and AKI in 421 (14.8%). Thirty-day mortality, ascertained via Social Security Death Index linkage, occurred in 31 patients (1.1%). Bacterial infections demonstrated significantly elevated indices versus viral: AISI (9,845 vs. 920, more than tenfold higher), NLR (14.8 vs. 1.82, more than eightfold higher; all p < 0.001). For bacterial-viral differentiation, NLR achieved AUC 0.892. In bacterial infections (n = 276), AISI demonstrated highest accuracy for AKI (AUC 0.896), ICU (0.882), and mortality (0.868); NPAR showed comparable results (AKI 0.854, ICU 0.872, mortality 0.848). In viral infections (n = 151), indices showed moderate performance: AISI for AKI (AUC 0.782), ICU (0.768), mortality (0.754); NPAR for AKI (0.764), ICU (0.756), mortality (0.742). HALP achieved strong mortality prediction in both groups (bacterial 0.874, viral 0.768). In diabetic patients (n = 530), AISI achieved strong discriminatory performance (AUC 0.908 for AKI) with higher complications (AKI 26.8% vs. 12.1%; p < 0.001). CKD patients (n = 296) showed highest complications (AKI 38.6%, mortality 2.1%). Multivariable analysis confirmed AISI (aOR 1.052), NPAR (aOR 1.046), and HALP (aOR 0.91) as independent predictors (all p < 0.001). Among nine CBC-derived indices evaluated in geriatric patients with acute infectious diarrhea, AISI, NPAR, and HALP demonstrate superior diagnostic value for predicting AKI, ICU admission, and mortality, with higher accuracy in bacterial compared to viral infections. These cost-effective biomarkers may address the current gap in geriatric-specific prognostic tools; however, pathogen-stratified AUC values pertain exclusively to microbiologically confirmed cases and represent upper-bound estimates requiring prospective validation with multiplex PCR-based pathogen ascertainment.