Comparison of coding data with clinical diagnosis of antibiotic-resistant healthcare-associated infections

Despite increasing use of International Classification of Disease (ICD) coding data to detect hospital-acquired complications, evidence of its accuracy is limited. Infections with multidrug resistant organisms (MROs) is a subcategory of healthcare-associated infection (HAI) with increased morbidity and mortality. This study compared ICD coding to clinical diagnosis for detection of MRO HAIs.

We calculated the positive predictive value (PPV) and estimated the sensitivity of hospital coding in identifying clinical MRO HAIs. We reviewed admissions in 2023 that had an MRO HAI code applied to assess for clinical evidence of MRO HAIs. The PPV of coded MRO HAI in detecting clinical MRO HAI was calculated. A second PPV, which excluded Staphylococcus aureus solely resistant to penicillin and Klebsiella species solely resistant to amoxicillin, was calculated to assess whether minor exclusions of expected resistance patterns of no clinical relevance could improve the PPV.

There were 262 MRO HAI codes across 122 relevant admissions in 2023. Clinical MRO HAI was confirmed in 76 admissions (PPV 62 %). Excluding expected resistance patterns increased the PPV to 73 %. Analysis of the 11 cases of healthcare-associated ESBL bacteraemia identified that three received an MRO HAI code (sensitivity 27 %).

Using coded data to determine presence of MRO HAIs is unlikely to be of sufficient accuracy to guide prevention activities or be used as a quality metric.