This retrospective single-center study evaluated whether a coding-based Pediatric Organ Dysfunction Index (PODI) derived from ICD-10 and OPS codes can approximate sepsis-related organ dysfunction and serve for risk adjustment in administrative healthcare datasets that lack detailed clinical and laboratory information. The authors used the pediatric Sequential Organ Failure Assessment (pSOFA) score as the reference standard and aimed to assess discrimination for in-hospital mortality, calibration, and organ-level agreement.
Pediatric sepsis episodes treated between November 2011 and November 2021 were identified at a single center. A total of 488 sepsis episodes were included. The observed in-hospital mortality in the cohort was 14.1%. The study received ethical approval from the Ethics committee of the Medical Faculty of the University of Duisburg Essen (22-10906-BO); informed consent was waived due to retrospective anonymized data collection. The authors report that study data are available upon reasonable request.
The PODI was constructed from routinely coded administrative data elements: International Classification of Diseases, 10th Revision (ICD-10) diagnosis codes and Operation and Procedure System (OPS) codes. The index was designed to capture sepsis-related organ dysfunction and therapy intensity where clinical and laboratory variables required for clinical organ dysfunction scores are not present in administrative datasets. An expanded version of the PODI that included additional ICD-10 and OPS codes was defined and evaluated in sensitivity analyses.
The pediatric Sequential Organ Failure Assessment (pSOFA) score served as the clinical reference standard. The authors compared the PODI to two pSOFA variants: pSOFA calculated at sepsis onset (pSOFAonset) and the maximum pSOFA during the episode (pSOFAmax). Agreement between PODI and pSOFA was quantified using Spearman's rank correlation for overall concordance and organ-specific analyses used sensitivity, specificity, positive predictive value, and negative predictive value.
For discriminating in-hospital mortality, the PODI demonstrated good performance with an area under the receiver operating characteristic curve (AUC) of 0.85 (95% CI 0.80–0.89). This discrimination was comparable to pSOFAmax (AUC 0.78, 95% CI 0.72–0.83) and superior to pSOFAonset (AUC 0.73, 95% CI 0.67–0.80). Calibration of the PODI was assessed in the study, though detailed calibration metrics were reported in the source and should be consulted there for specifics.
Agreement between the coding-based PODI and clinical pSOFA components varied substantially by organ system. Sensitivity to detect pulmonary dysfunction was highest among organ-specific comparisons, while other organ systems showed lower sensitivity or variable predictive values. Overall Spearman correlations were moderate: 0.54 when comparing PODI to pSOFAonset and 0.60 when comparing to pSOFAmax. The authors emphasize that similar predictive performance for mortality does not imply that the two scores are interchangeable at the individual-patient level.
An expanded PODI that incorporated additional ICD-10 and OPS codes increased organ-level sensitivity for selected components, indicating that adding codes can improve capture of some dysfunctions from administrative records. However, this expansion did not meaningfully improve discrimination for in-hospital mortality beyond the standard PODI. Specific organ-level improvements and the list of additional codes are provided in the study supplementary materials.
The authors propose that the standard coding-based PODI may be a practical covariate for adjusting for organ dysfunction and therapy intensity in large administrative datasets that use ICD-10 coding, where clinical and laboratory data are not available. They caution that the PODI showed only moderate agreement with the clinical pSOFA, and therefore it should be applied for risk adjustment at the group level rather than as a substitute for clinical organ dysfunction scores in individual patients. Limitations include single-center retrospective design and the need for broader validation. The manuscript notes the need for further validation and refinement in non-sepsis cohorts before recommending widespread implementation in large-scale administrative research.
In this cohort of 488 pediatric sepsis episodes, a coding-based PODI discriminated in-hospital mortality well and may offer a feasible method for risk adjustment when clinical organ dysfunction scores cannot be derived from administrative data. Moderate correlation with pSOFA and variable organ-level agreement indicate the PODI should not replace clinical scoring for individual patients. The authors recommend additional validation and refinement of the index, including testing in other centers and non-sepsis populations, before broader use in administrative-data research.