Professional Documents
Culture Documents
DOI 10.1007/s00228-016-2026-0
Received: 29 September 2015 / Accepted: 10 February 2016 / Published online: 20 February 2016
# Springer-Verlag Berlin Heidelberg 2016
Abstract Results Out of the 1388 adult deaths studied, 256 (18.4 %)
Purpose Hospital mortality related to adverse drug reactions were suspected of being related to drugs. Drugs were
(ADRs) is a relevant clinical problem with major health and suspected of causing death in 146 inpatients (10.5 %) and
economic consequences. We conducted a study to assess hospi- contributing to death in 110 (7.9 %). Drugs related to death
tal mortality related to ADRs, the drugs most frequently in- were administered during the hospital stay in 161 cases
volved, and the possible risk factors associated with fatal ADRs. (11.5 %) and before hospital admission in 95 (6.84 %). The
Methods A retrospective observational study was conducted, most frequent fatal ADRs were cardiac arrhythmia, gastroin-
reviewing the clinical records of 1388 consecutive adult pa- testinal bleeding, and respiratory failure. The drugs most fre-
tients (18–101 years) who died during a 22-month period in a quently involved in fatal ADRs were antithrombotics (antico-
tertiary hospital in Southern Europe (Granada, Spain). The agulants or antiplatelets) (23 %), psychotropic drugs (21.2 %),
main outcome was the prevalence of hospital death suspected and digoxin (11.3 %). Independent risk factors for ADR-
to be related to administered drugs. related death were the presence of ≥4 diseases (OR = 1.43)
and the receipt of ≥10 drugs (OR = 3.24), but no significant
Electronic supplementary material The online version of this article association with gender or age was found.
(doi:10.1007/s00228-016-2026-0) contains supplementary material, Conclusions A high percentage of hospital deaths were
which is available to authorized users. suspected of being associated with ADRs, especially in patients
with comorbidity and/or polypharmacy. Antithrombotics, psy-
* Alfredo José Pardo Cabello chotropics, and digoxin were the drugs most frequently associ-
apardoc05@yahoo.es
ated with in-hospital drug-related deaths.
1
Department of Internal Medicine, San Cecilio University Hospital, Keywords Adverse drug reactions . Hospital mortality .
Avda. Dr. Olóriz n°16, 18012 Granada, Spain Inpatients . Epidemiology
2
Department of Pharmacology, School of Medicine, Biomedical
Research Institute ibs. Granada, University of Granada,
Granada, Spain
3
Department of Medicine, School of Medicine, University of Introduction
Granada, Granada, Spain
4
School of Health Sciences of the University of Granada, Drugs are widely used in the control of disease currently,
Granada, Spain but it is well known that they are not free from risk of
5
Department of Biostatistics, School of Medicine, University of harm, especially in some clinical circumstances. The ad-
Granada, Granada, Spain verse drug reactions associated to pharmacological treat-
6
Clinical Pharmacology Unit of San Cecilio University Hospital, ments remains an unresolved clinical challenge with
Granada, Spain health and economic implications, which affects patients
7
Department of Pharmacology of School of Medicine, University of in hospital and community settings [1–5]. An extensive
Granada, Granada, Spain literature review on drug-related deaths in hospital
732 Eur J Clin Pharmacol (2016) 72:731–736
patients reported a wide range of prevalences (0.30–19 %) out of the 1388 studied. When the researchers could not reach
depending on the methodology employed [6–17]. a unanimous agreement after the discussion, the association of
A previous meta-analysis reported that adverse drug reac- drug(s) with the death in question was rejected. The final
tions (ADRs) affected 13 % of adult patients in Spanish hos- decision was entered in a database with the aforementioned
pitals and were strongly associated with polypharmacy and study variables. The anonymity of the patients and the confi-
advanced age, being the cause of death in 0.1 % of hospital- dentiality of their data were preserved at all times, in accor-
ized patients [18]. A pilot observational study of 289 adult dance with Spanish data protection legislation. The committee
patients dying during their stay in our hospital revealed that comprised two internal medicine specialists (F.G.J. and
5.9 % of the deaths were related to their pharmacological A.P.C.), two clinical pharmacologists (E.P.G. and E.P.C.),
treatment according to WHO methodology [19]. Based on and a nurse (C.M.R.), all with wide experience in the clinical
these preliminary results, a wider study was designed to assess setting and in the detection of ADR in inpatients. The data
the prevalence of drug-related death and the associated risk obtained were previously checked and then entered into a
factors in a larger sample of patients dying in the hospital SPSS 20.0 database for statistical analysis.
during a 22-month period.
Identification of fatal ADRs and causality assessment
Table 2 Logistic regression analysis of risk factors associated with In the present study, the drugs most frequently suspected to
ADRs-related deaths in the hospital
be associated to death were antithrombotics, psychotropics,
Non-adjusted model Adjusted model and digoxin, in line with previous reports [10, 19]. Fatal
ADR-related suspected cases in our study included the follow-
Variable OR (95 % CI) OR (95 % CI) ing: sudden death [26] and bronchoaspiration [27] associated
Sex with consumption of single or multiple psychotropic drugs
Male 1.00 1.00 [28]; heart failure due to digitalis toxicity, mainly in females,
Female 1.14 (0.87–1.49) 1.12 (0.84–1.49) as also observed in a study by the digitalis investigation group
Age (DIG) [29]; acute respiratory failure in patients with a history
18–68 1.00 1.00 of respiratory insufficiency after taking opiates alone or in
69–78 1.36 (0.91–2.04) 1.19 (0.77–1.85) association with sedatives [30]; severe gastrointestinal and
79–84 1.94 (1.3–2.9) 1.45 (0.92–2.3) cerebral hemorrhages in patients receiving single or multiple
>84 1.61 (1.07–2.41) 1.29 (0.81–2.06) antithrombotics alone or associated with NSAIDs or corti-
Number of diseases coids [10, 16, 19]; and hip fractures or traumatic brain injuries
≤4 1.00 1.00 from falls at home in elderly patients receiving psychotropic
>4 1.82 (1.38–2.4) 1.43 (1.04–1.95) or antihypertensive drugs [31], causing their admission to the
Number of drugs emergency department (Supplementary Material 1). Most of
0–5 1.00 1.00 the drugs associated with fatal ADRs are commonly used to
6–9 1.53 (0.95–2.45) 1.46 (0.9–2.39) treat pain and cardiovascular, respiratory, or psychiatric disor-
10–12 3.53 (2.29–5.45) 3.24 (2.05–5.12) ders, the most prevalent diseases among inpatients.
>12 4.94 (3.17–7.7) 4.44 (2.77–7.11) It appears that a high proportion of these cases might po-
Department tentially be prevented, underlining the need to prescribe these
Oncology 1.00 1.00 drugs with special care in elderly patients with multiple dis-
Intensive Care 1.48 (0.83–2.64) 0.71 (0.38–1.35) eases; however, the preventability of fatal ADRs is a difficult
Internal Medicinea 1.81 (1.13–2.9) 0.79 (0.45–1.4) issue, especially in retrospective studies [32]. We also found
Surgeryb 2.75 (1.61–4.7) 1.32 (0.71–2.44) that a third of drug-related deaths were associated with drug-
Length of stay 1.02 (1–1.03) 1.01 (0.99–1.02) drug interactions, largely pharmacodynamic, consistent with
the high prevalence (28 %) of interactions reported for all
ADRs adverse drug reactions, OR odds ratio, CI confidence Interval medical prescriptions in inpatients [33].
a
And other medical specialties The number of drugs received by patients and the number
b
And other surgical specialties of their diseases emerged as independent risk factors for
ADR-related death in our study, as indicated by the multilevel
adjusted regression analysis. The risk of fatal ADR has previ-
of the drugs responsible is more challenging in dead patients. ously been associated with polypharmacy [9, 14, 34], but not
Our study showed an elevated prevalence (18.4 %) of hospital all studies have found an association with the number of as-
deaths suspected of being caused by or contributed to by sociated diseases [9, 14, 19]. Age and admission in a surgery
drugs. This prevalence is very similar to the percentage report- department did not behave as independent risk factors for
ed by Ebbesen et al. [9] in a Norwegian Internal Medicine drug-related death, with the odds ratio for these variables be-
Department (18.2 %) using the present methodology. These ing markedly reduced and losing statistical significance when
prevalences are markedly higher than those of 5–6.4 % report- the analysis was adjusted for the number of drugs and
ed by other researchers using a different approach [16, 19], diseases.
which may explain the discrepancy [23, 24]. Our findings may Study strengths include the wide sample, which was repre-
reflect the greater sensitivity of the present method to detect sentative of all hospitalized patients in our setting, and the
treatment-related deaths in comparison to techniques based on application of a previously validated methodology. A further
algorithms and continuous scales like Naranjo’s one. This is strength of our approach was the case-by-case classification of
because of the lack of information on the response of the deaths as based on the unanimous decision of a multidisciplin-
individual to the withdrawal and re-administration of the sus- ary committee after an exhaustive and independent review of
pect drug, which are both considered crucial for the definite all cases of death. This procedure contrasts with studies based
imputation of a drug using these methods [25]. Though solely on administrative data, which are considered to sub-
Naranjo’s algorithm is the most frequently used, a previous stantially underestimate the percentage of ADRs in admitted
study by our group [19] found that this scale had a lower patients, whether fatal or not [24, 35, 36].
sensitivity to detect fatal ADR in deceased patients with re- This study has some limitations: (i) the retrospective nature
spect to the method [9, 20] used in the present study. of the study may underestimate the occurrence of ADRs if
Eur J Clin Pharmacol (2016) 72:731–736 735