About on-sixth of inpatients were found to have experienced ADEs, at a mean rate of 18.3 ADEs per 100 patients. A meta-analysis of observational studies presented estimates of ADE according to the method of identification of events . In two out of twenty five studies the events were identified by a similar set of triggers. One of them was conducted in the United States with six community hospitals and estimated a rate of 15.0 ADEs per 100 patients . In the other study, researchers encountered in a Brazilian hospital a rate of 26.6 ADEs per 100 patients . Although the event rate in the last study is higher than ours the proportion of people with ADE estimated by the authors  is similar, around 15%. Some characteristics of the study population and the hospital’s profile may explain the differences. The differences in rates may be also attributed to staff education or case mix that may occur even when comparing data from a single country.
The characteristics of ADEs we focused on were patient symptoms and degree of harm. Rashes, nausea, vomiting, pruritus and dizziness accounted for almost half the events. The profile of ADEs identified in our study is similar to that identified in a study conducted in a tertiary care hospital in Northern Brazil, where skin was found to be the most commonly affected organ system. The gastrointestinal system was also among the three most affected of them .
As regards degree of harm, most of the events resulted in temporary patient harm that required some intervention. Other studies also used the same source to classify ADEs, which is the National Coordinating Council for Medication Error Reporting and Prevention Index. They obtained similar proportions of events of lower degree of severity, 87%  and 79.9% . Life-threatening events were much less common, as in other studies . We also identified events that required intervention to keep the patient alive, such as hypoglycemia, cardiac tamponade and over-sedation, related, respectively, to the use of insulin, heparin and midazolam.
In our sample, certain classes of drugs are intensively prescribed and also strongly associated with ADEs (i.e., return higher ratios of “related ADEs” to “number of prescriptions”), they were analgesics, antibacterials for systemic use, antithrombotic agents and drugs for acid-related disorders. Ranitidine is a drug from this last subclass, as it is a histamine H2 receptor antagonists, and serves to illustrate the problem.
Ranitidine was implicated in cases of rash. It was the most frequent event identified in our study. It was prescribed for 70% of patients, a value compatible with those found in the literature for use in therapy to suppress gastric acid production [27, 28]. We observed that, in most patients, ranitidine was being used as prophylactic medication rather than to treat gastrointestinal diseases. Stress ulcer prophylaxis has become an increasingly common practice for clinical patients, although there is little or no clinical evidence to support it  and it can be considered unnecessary in 73% of cases . According to figures from this hospital, more than 50% of the ranitidine dispensed is for intravenous administration, which exposes the patients to unnecessary risk, because the injection route can cause local burning, itching, skin rash and vasculitis .
The occurrence of ADEs is associated with the number of drugs used [5, 25, 29–31]. Our data corroborate that association: the likelihood of an ADE occurring was higher in users of 10 or more drugs than in those who used fewer drugs during their hospital stay (p value <0.001), although the analysis was not adjusted for confounders. When patients are seriously ill, it is often difficult to evaluate the degree of harm to be attributed to the number of prescribed drugs or to drug classes. In our study population, one patient suffered a bed fall during the hospital stay. Besides the fact that he was exposed to drugs likely to cause falls (captopril, hydrochlorothiazide), evaluation of the harm was jeopardised by the complexity of the patient’s condition (serious traffic accident casualty).
Patients with longer hospital stays are exposed to greater likelihood of ADE than those hospitalised for up to nine days (p value <0.001). This finding is consistent with those of several previous studies . Hence it is important draw attention to long-stay inpatients, because in addition to being prone to ADEs, the tendency is towards multiple events and more pronounced patient harm.
As longer hospital stays and use of more drugs are regarded as factors associated with ADEs, our findings may help physicians to identify at-risk patients and monitor them carefully. Moreover, to prevent ADES during hospital stay, the risk factors can be categorised according to opportunities for intervention, such as decreasing the number of drugs prescribed or adjusting the dose .
This study has limitations in terms of the internal validity. The results of specific events should be analysed with caution because the sample size was calculated considering the estimated global incidence of ADEs. Thus, the frequency of specific events may reflect random variations and may be skewed.
Although the study is not large, we did succeed in observing values of the estimates that differed with statistical significance between patients with or without ADEs for the variables length of stay and number of medications in the bivariate analysis. However studies with larger samples should test the hypothesis of association.
The study was conducted at only one hospital, which makes external validity troublesome and places limitations on how far the results can be generalised. Nonetheless, they may be applied to other tertiary hospitals in medical schools in Brazil and in other countries.
We endeavoured to increase internal validity by using double data collection and assembling a team of researchers and health professionals to evaluate the cases. In order to improve objectivity, the method was applied at all stages by two independent reviewers, using a specific manual containing definitions and detailed descriptions of the procedures. Nonetheless, some subjectivity in identifying and interpreting triggers and events cannot be ruled out, as demonstrated in other retrospective patient record review studies . The double-blind evaluation performed here was not intended to assess inter-observer reliability, but rather to foster information completeness and improve validity of the information. The double evaluation process with two physicians of patient records to assess ADE is not more reliable than a record review process with one physician .
Poor quality of information in hospital records is a common problem in chart review-based research , but the choice of a teaching hospital possibly reduces information bias. There were concerns about data completeness, however. Conspicuous gaps in the recorded information included race/ethnicity and occupation, which were missing from about 20% of patient records.
During application of the method, two events unrelated to the triggers were identified, one of them a case of increased blood transaminase levels resulting from use of phenytoin. Studies have shown that laboratory test results for transaminases levels can be effective triggers for detecting ADEs. Aspartate aminotransferase (AST) screening returns a positive predictive value between 0.01 and 0.23, and alanine aminotransferase (ALT), a value from 0 to 0.31 . The fact that, in our study, some ADEs were not found by triggers reinforces the idea that no single method captures every event that occurs in hospitals. This challenges managers and policy makers to test and harmonise different methods to address adverse events. The use of triggers to identify ADEs, when integrated with event monitoring, stimulated or unstimulated spontaneous reporting and other techniques, can be an important strategy for indicating possible shortcomings in the process of using medications for hospitalised patients.
According to Rozich  the trigger enables organisations to monitor longitudinally how ADE rates change in response to strategies designed to improve clinical safety . However, this strategy is nowadays mainly theoretical, since there are several obstacles to be overcome, among them: scant evidence of impact of ongoing strategies; the existence of different subcategories of drug adverse events requiring different improvement interventions; the fact that many common adverse events require more complex detection strategies [37, 38].
The concept of ADE we used included drug adverse reactions (ADR) and errors. It does not conflict with the concept of ADR that sustains the algorithm used to determine the strength of the causal relationship with the drugs in our study and others [39–41]. Steady conceptual advances in the field of pharmacoepidemiology are permitting studies focussing primarily on occurrences of patient harm, regardless of whether or not such harm is associated with errors in therapeutic indication or drug administration. Besides, examining carefully the patient charts we did not identified any case of error like non adherence and subtherapeutic doses.
In 2013 Brazil’s Ministry of Health launched a national patient safety programme . It hinges on engaging patients in health care, including the subject of ‘patient safety’ in undergraduate and postgraduate education, and increasing research. On this latter item, the programme specifies measuring harm, understanding the causes, identifying the solutions, assessing the impact and applying the evidence to assure safer patient care. Given that framework, we believe the results of this study can help to call attention to the need of develop researchs aiming to estimate and characterize ADE and medication involved as well as the role of number of drugs prescribed for inpatients.