Out-of-Hospital Use of PPIs and Hypomagnesemia
A flow diagram describing the derivation of the study population is shown in Fig 1. In brief, during the 7-year study period, there were 97,472 hospitalizations representing 51,207 patients. Hospitalizations for surgical, obstetric, psychiatric, oncologic, and substance abuse services were excluded (31,353 hospitalizations; 17,027 patients). Other reasons for exclusion were the absence of a serum magnesium value on the day of admission (20,522 hospitalizations; 7,802 patients) and missing service assignment (4,994 hospitalizations; 2,798 patients). By limiting to the first hospitalization per patient, 17,023 hospitalizations were excluded. Of the 23,580 patients with a first hospitalization on medical services, Cage was 70 years, and 40% of patients were men. Comorbid conditions were more prevalent in cases compared with controls, as evidenced by higher Charlson-Deyo comorbidity index scores (global P = 0.04) and a higher prevalence of diabetes (P = 0.01). Cases also had a higher prevalence of out-of-hospital thiazide diuretic use (P = 0.02) and lower serum calcium levels (P < 0.001) compared with controls. GERD unexpectedly was more prevalent in controls compared with cases (P = 0.03).
(Enlarge Image)
Figure 1.
Flow Diagram Describes the Derivation of the Analytical Data Set.
Table 1 lists characteristics of study participants. Mean age was 70 years, and 40% of patients were men. Comorbid conditions were more prevalent in cases compared with controls, as evidenced by higher Charlson-Deyo comorbidity index scores (global P = 0.04) and a higher prevalence of diabetes (P = 0.01). Cases also had a higher prevalence of out-of-hospital thiazide diuretic use (P = 0.02) and lower serum calcium levels (P < 0.001) compared with controls. GERD unexpectedly was more prevalent in controls compared with cases (P = 0.03).
Based on a manual chart review of 239 hospitalizations as the gold standard, the diagnostic performance of method 1 to ascertain PPI use (review of physician admission and discharge notes in the electronic medical record) was better than that of method 2 (review of nursing electronic medication administration record), as listed in Table 2.
In unadjusted and adjusted analyses (Table 3), out-of-hospital PPI use was not associated with higher odds for a low serum magnesium level at the time of hospital admission (adjusted OR [aOR], 0.82; 95% CI, 0.61–1.11). There was no significant interaction between PPI use and each covariate (data not shown). Similarly, there was no association between type of PPI (global P = 0.1) or the omeprazole equivalent daily dose (aOR, 0.98; 95% CI, 0.88–1.10) and serum magnesium levels. Other factors independently associated with low serum magnesium levels were diabetes (aOR, 1.42; 95% CI, 1.01–1.98; P = 0.04) and out-of-hospital thiazide diuretic use (aOR, 1.95; 95% CI, 1.18–3.22; P = 0.009; Table S3).
There was no association between PPI use and low serum magnesium levels in sensitivity analyses restricted to case-control pairs in which ascertainment of PPI use was based on method 1 that had better diagnostic performance, patients with severe hypomagnesemia, those with preserved kidney function, or esophageal disorders (Table 3). Although out-of-hospital PPI use was associated with a 1.16 higher adjusted odds for low serum magnesium levels in an analysis restricted to patients with GERD, this association did not reach statistical significance (OR, 1.16; 95% CI, 0.76–1.76).
We identified and removed 7 potentially influential observations from our primary multivariable model, which consisted of cases (with 6 women) and their matched-pair control. The aOR for PPI use without these observations was approximately the same (OR, 0.76; 95% CI, 0.56–1.03) as when using the full data set (OR, 0.82; 95% CI, 0.61–1.11).
Results
Derivation of the Study Population
A flow diagram describing the derivation of the study population is shown in Fig 1. In brief, during the 7-year study period, there were 97,472 hospitalizations representing 51,207 patients. Hospitalizations for surgical, obstetric, psychiatric, oncologic, and substance abuse services were excluded (31,353 hospitalizations; 17,027 patients). Other reasons for exclusion were the absence of a serum magnesium value on the day of admission (20,522 hospitalizations; 7,802 patients) and missing service assignment (4,994 hospitalizations; 2,798 patients). By limiting to the first hospitalization per patient, 17,023 hospitalizations were excluded. Of the 23,580 patients with a first hospitalization on medical services, Cage was 70 years, and 40% of patients were men. Comorbid conditions were more prevalent in cases compared with controls, as evidenced by higher Charlson-Deyo comorbidity index scores (global P = 0.04) and a higher prevalence of diabetes (P = 0.01). Cases also had a higher prevalence of out-of-hospital thiazide diuretic use (P = 0.02) and lower serum calcium levels (P < 0.001) compared with controls. GERD unexpectedly was more prevalent in controls compared with cases (P = 0.03).
(Enlarge Image)
Figure 1.
Flow Diagram Describes the Derivation of the Analytical Data Set.
Population Characteristics
Table 1 lists characteristics of study participants. Mean age was 70 years, and 40% of patients were men. Comorbid conditions were more prevalent in cases compared with controls, as evidenced by higher Charlson-Deyo comorbidity index scores (global P = 0.04) and a higher prevalence of diabetes (P = 0.01). Cases also had a higher prevalence of out-of-hospital thiazide diuretic use (P = 0.02) and lower serum calcium levels (P < 0.001) compared with controls. GERD unexpectedly was more prevalent in controls compared with cases (P = 0.03).
Diagnostic Performance of PPI Use Ascertainment Methods
Based on a manual chart review of 239 hospitalizations as the gold standard, the diagnostic performance of method 1 to ascertain PPI use (review of physician admission and discharge notes in the electronic medical record) was better than that of method 2 (review of nursing electronic medication administration record), as listed in Table 2.
Primary Analyses
In unadjusted and adjusted analyses (Table 3), out-of-hospital PPI use was not associated with higher odds for a low serum magnesium level at the time of hospital admission (adjusted OR [aOR], 0.82; 95% CI, 0.61–1.11). There was no significant interaction between PPI use and each covariate (data not shown). Similarly, there was no association between type of PPI (global P = 0.1) or the omeprazole equivalent daily dose (aOR, 0.98; 95% CI, 0.88–1.10) and serum magnesium levels. Other factors independently associated with low serum magnesium levels were diabetes (aOR, 1.42; 95% CI, 1.01–1.98; P = 0.04) and out-of-hospital thiazide diuretic use (aOR, 1.95; 95% CI, 1.18–3.22; P = 0.009; Table S3).
Sensitivity Analyses
There was no association between PPI use and low serum magnesium levels in sensitivity analyses restricted to case-control pairs in which ascertainment of PPI use was based on method 1 that had better diagnostic performance, patients with severe hypomagnesemia, those with preserved kidney function, or esophageal disorders (Table 3). Although out-of-hospital PPI use was associated with a 1.16 higher adjusted odds for low serum magnesium levels in an analysis restricted to patients with GERD, this association did not reach statistical significance (OR, 1.16; 95% CI, 0.76–1.76).
We identified and removed 7 potentially influential observations from our primary multivariable model, which consisted of cases (with 6 women) and their matched-pair control. The aOR for PPI use without these observations was approximately the same (OR, 0.76; 95% CI, 0.56–1.03) as when using the full data set (OR, 0.82; 95% CI, 0.61–1.11).
Source...