(184) Validity of Inpatient ICD 10 Codes for Heat-related Illnesses in US Medicare Data

Background: Heat-related illnesses (HRI) represent a significant yet underrecognized cause of hospitalization, particularly in the context of global climate change. The validity of International Classification of Diseases, Tenth Revision (ICD-10) codes in identifying hospitalizations due to HRI in claims data is unknown.

Objectives: To determine the validity of Inpatient ICD 10 Codes for Heat-related Illnesses in US Medicare Data

Methods: We used 20% of national Medicare data from 2015-2019 to assess the validity of ICD-10 codes (T67.0 heat stroke and T67.1/.2 heat syncope) in identifying hospitalizations attributed to heat-related illnesses. We randomly sampled and obtained medical records (MRs) for hospital admissions with primary diagnoses of the T67 codes. The gold standard (GS) for a heat-related hospitalization was a documented diagnosis of heat stroke or heat syncope by a treating physician (GS1) or adjudicated HRI diagnosis confirmed by documented clinical signs of hyperthermia, dehydration, or heat-related syndromes (GS2). Data were extracted from MRs by a trained health professional using a standardized form and reviewed by certified physicians. The positive predictive values (PPVs) with 95% confidence intervals (CIs) were calculated to evaluate the validity.

Results: Of the 153 cases reviewed (mean age 73, 59% male, and 61% White), 80 were coded as heat stroke (T67.0), and 73 were heat syncope (T67.1/.2). The PPV for heat stroke was 73% [95% CI, 63-82] for GS1 and 62% [95% CI, 52-73] for GS2. The PPV for heat syncope was 65% [95% CI, 54-76] for GS1 and 68% [95% CI, 57-78] for GS2. When diagnoses for other HRI, such as heat exhaustion, were included as a part of GS1 or GS2, PPV for heat stroke and syncope improved by 5-7% (PPV 67-78%). Detailed clinical documentation supporting the diagnosis varied and was absent in many cases. Body temperature was absent in 3% but taken at different times and locations, including the emergency room and by paramedics on-site, with some temperatures recorded post-ice pack administration. Heat exposure status was recorded in 79% of the records. Patient factors, including hydration status and physical activity, and environmental factors, such as occupation and living conditions, were often missing in the records.

Conclusions: Heat stroke can be identified using ICD-10 codes with moderate to high accuracy in Medicare data. However, the variation in PPVs for other heat-related conditions suggests the need for improved clinical documentation and potential refinement of ICD-10 coding practices to better capture the spectrum of HRI. These findings underscore the importance of clinical awareness and active surveillance to mitigate the impact of heat-related illnesses, especially among the elderly and other high-risk groups.