Analysis of electronic health records from three distinct and large populations reveals high prevalence and biases in the co-administration of drugs known to interact

Abstract

The co-administration of drugs known to interact has a high impact on morbidity, mortality, and health economics. We study the drug-drug interaction (DDI) phenomenon by analyzing drug administrations from population-wide Electronic Health Records (EHR) in Blumenau (Brazil), Catalonia (Spain), and Indianapolis (USA). Despite very different health care systems and drug availability, we find a common large risk of DDI administration that affected 13 to 20% of patients in these populations. In addition, the increasing risk of DDI as patients age is very similar across all three populations but is not explained solely by higher co-administration rates in the elderly. We also find that women are at higher risk of DDI overall- except for men over 50 years old in Indianapolis. Finally, we show that PPI alternatives to Omeprazole can reduce the number of patients affected by known DDIs by up to 21% in both Blumenau and Catalonia, and 2% in Indianapolis, exemplifying how analysis of EHR data can lead to a significant reduction of DDI and its associated human and economic costs. Although the risk of DDIs increases with age, administration patterns point to a complex phenomenon that cannot be solely explained by polypharmacy and multimorbidity. The lack of safer drug alternatives, particularly for chronic conditions, further overburdens health systems, thus highlighting the need for disruptive drug research.

Keywords: drug-drug interactions; electronic health records; multimorbidity.

Figure 1.

Risk of co-administration and interaction by age during the first 18 months of the studies. Green, red, and blue lines denote measurements for Blumenau, Catalonia, and Indianapolis, respectively. (a) Risk of co-administration of drugs, RC^[y1,y2] (eq. 10). (b) Risk of co-administration of drugs known to interact, RI^[y1,y2] (eq. 11). (c-e) RI^[y1,y2] against respective null model ${\widehat{RI}}^{[y1,y2]}$ in (c) Blumenau, (d) Catalonia, and (e) Indianapolis. Circles denote the values obtained with the real data, while the asterisks denote the values obtained using the null model. The associated relative risk (eq. 13) is shown above the points. Asterisks denote significant differences (Fisher’s exact test).

Figure 2.

(a-c) Risk of drug co-administration, RC^[y1,y2],g, and (d-f) interaction, RI^[y1,y2],g, by age and sex (as defined in section 4.6) for Blumenau, Catalonia, and Indianapolis in the first 18 months of administration. Red and blue colors denote the risks in women (g = W) and men (g = M), respectively. Relative risks of co-administration (RRC^[y1,y2],W) and interaction (RRI^[y1,y2],W) for women per age group displayed above the points (as defined in section 4.5 and section 4.6). Asterisks denote significant differences (Fisher’s exact test).

Figure 3.

Catalonia DDI Network. Nodes denote drugs i involved in at least one co-administration known to be a DDI. Only nodes connected via edges with ${\tau }_{i,j}^{\varphi }>0.18$ are shown for clarity. Node color represents the highest level of primary action class, as retrieved from drugs.com. Node size proportional to P(U_i^Φ) per eq. 12, the probability of patients being affected by a DDI involving drug i. Edge weights denote strength of interaction, τ_i,j^Φ per eq. 5. Edge colors denote $RR{I}_{i,j}^g$, where g ∈ M, W, to identify DDI edges that are higher risk for women (red) or men (blue). Color intensity for $RR{I}_{i,j}^g$ varies in [1,5]; that is, values are clipped at 5 for clarity.

Figure 4.

Top 20 drug interactions with the highest difference between RI^[y1,y2],W and RI^[y1,y2],M (see eq. 11). Colors denote a higher risk of interaction for women (red) and men (blue). Markers (+ and −) denote significantly higher risk of DDI administration for the respective sex after correcting for multiple testing (FDR ≤ 0.05). Note color scale is different across populations, as the maximum and minimum differences in RI^[y1,y2] are different between populations.

Figure 5.

Diagram of co-administration and interaction computation for Catalonia, Blumenau, and Indianapolis. Two hypothetical patient-drug dispensing timelines with three drugs (i, j, & k) are represented. In Catalonia (left), two drugs (i,j) are assumed to be co-administered if they were dispensed and billed during the same month. In Blumenau and Indianapolis (right), two drugs are assumed to be co-administered if they were dispensed for an administration period with an overlap of at least one day. Drug administration lengths (in days for Blumenau and Indianapolis, and months for Catalonia) are shown for each dispensation. The three possible pairwise comparisons (i,j). (i,k), and (j,k) between the dispensed drugs are shown with their co-administration overlap marked with backgrounds in either orange (not known DDI) or red (known DDI).