Fig. 1

The association between antipsychotic use and hospitalization with hyponatremia assessed in four subgroups^a. CI confidence interval. a Antipsychotic type, antipsychotic dose, chronic kidney disease, congestive heart failure, and use of a diuretic. Sets of medication users and non-users were matched on presence of chronic kidney disease, congestive heart failure, and baseline diuretic use. For antipsychotic type and dose, matched sets were categorized according to this characteristic in users. b Hyponatremia (and the proportion of patients who had an event) was assessed by using a hospital diagnosis code. The true event rate of hyponatremia is underestimated for some outcomes because the code for hyponatremia has high specificity but low sensitivity. c Higher dose was defined as a higher than median daily dose. See Additional file 1: Table S2 for definitions. d Congestive heart failure has a sensitivity, specificity, and positive predictive value of 84.3, 85.4, and 35.8 %, respectively [31]. e Diuretic use includes potassium sparing and non-potassium sparing medications. Chronic kidney disease was included in the test for interactions but was removed from presentation as there were too few events for meaningful analysis. This was also done to comply with privacy regulations, to prevent the risk of re-identification when the size of the numerator is small (less than or equal to 5). Chronic kidney disease was identified by using an algorithm of hospital diagnosis codes validated for older adults in the study region [30]. The algorithm identified patients with a median estimated glomerular filtration rate of 38 mL/min/1.73 m² (interquartile range, 27–52 mL/min/1.73 m²), whereas its absence identified patients with a median estimated glomerular filtration rate of 69 mL/min/1.73 m² (interquartile range, 56–82 mL/min/1.73 m²). Data marker size is proportional to the inverse of the source variance