Abstract
Background and Aims: The positive predictive value (PPV) of a hyperkalemia diagnosis in administrative data has not been evaluated and relationships between coded hyperkalemia diagnoses, elevated potassium concentrations and clinical signs have not been assessed. The aims of this study were to determine the performance of a coded diagnosis in administrative data at identifying 1) clinically-evident hyperkalemia and 2) patients with potassium >= 6 mmol/liter.
Methods: This retrospective observational study included 8,722 patients with diabetes newly-initiating ACEi, ARB, or spironolactone therapy. The outcome measure was the first hyperkalemia-associated event comprised of a hospitalization, emergency department visit or death that occurred within 24 hours of a coded hyperkalemia diagnosis and/or potassium >= 6 mmol/liter during the first year of therapy. The medical records of all patients identified from administrative data as having hyperkalemia-associated events were reviewed as were a random sample of records of patients identified as not having hyperkalemia.
Results: Among the random sample (n = 99) of patients identified from administrative data as not having hyperkalemia, none had hyperkalemia upon record review. Among 64 patients identified from administrative data as having hyperkalemia, all had a hospitalization or emergency department visit associated with a coded hyperkalemia diagnosis (n = 41), potassium >= 6 (n = 9), or coded diagnosis and potassium >= 6 (n = 14). Of the 55 total patients with a coded hyperkalemia diagnosis, 42 (PPV 76%) had clinically-evident hyperkalemia; 32 (PPV 58%) had potassium >= 6. Of the 9 patients identified with potassium >= 6 who had no coded hyperkalemia diagnosis, 7 (PPV 78%) had clinically-evident hyperkalemia. Although 5 of the 64 patients died at the time of the hyperkalemia outcome, none had a coded diagnosis of hyperkalemia-associated death and hyperkalemia was not mentioned on the death certificate of any patient.
Conclusions: Nearly one-fourth of patients with a coded hyperkalemia diagnosis do not have clinical signs and nearly one-half do not have potassium concentration >= 6 mmol/liter. Because both false positives and negatives occur when using coded hyperkalemia diagnoses to identify hyperkalemia-associated outcomes from administrative and death certificate data, medical record validation of hyperkalemia outcomes is necessary.
- Received May 27, 2010.
- Accepted May 27, 2010.
