Performance of triggers in detecting hospitalizations related to drug-induced respiratory disorders in older adults: A pilot cross-sectional study

Highlights • A prevalence of 8.3 % of drug-induced respiratory disorders was identified.• Triggers detected hospitalizations related to drug-induced respiratory disorders.• Four triggers showed good performance for detecting drug-induced respiratory disorder.• Two triggers detected therapeutic ineffectiveness related to respiratory symptoms.


Introduction
Considering the need for care and interventions within the scope of health, especially after the COVID-19 pandemic, there has been an exponential increase in studies evaluating respiratory disorders and their risk factors. 1 However, there are still gaps in the literature regarding drug-induced respiratory disorders, which can be responsible for more than one respiratory disease and affect the entire respiratory system (e.g., airways and respiratory muscles). 2 drug-induced respiratory disorder is a type of Adverse Drug Event (ADE) with high morbidity and mortality 3 and is characterized by the presence of nonspecific clinical signs and symptoms common to respiratory diseases (e.g., cough, dyspnea, and fever).3,4 Although it is known that older people, particularly those with a previous diagnosis of respiratory diseases or infections, 3 who are receiving multiple drugs (polypharmacy), 5 or using analgesics and antibiotics, 6 are at greater risk of developing drug-induced respiratory disorders, 2 there is a lack of studies designed to detect this type of ADEs and related hospitalizations.In addition, there is no consensus in the literature on whether drug-induced respiratory disorders have a low prevalence or if they are under-detected, with an estimated prevalence in older people around 7.5 %. 6 Detecting drug-induced respiratory disorders is challenging and typically involves excluding other possible causes.3 In the absence of a gold-standard method, the application of triggers can be a valid strategy to detect and quantify the prevalence of drug-induced disorders, as they have shown good performance and utility in detecting ADEs and determining their preventability in health care settings.7 Furthermore, the easy application and reproducibility of triggers in different health services contribute to the planning and implementation of health interventions, thereby enhancing patient safety. 7In previous studies, triggers have demonstrated good performance in detecting ADEs, such as delirium and constipation in older people.7,8 Nonetheless, to the best of our knowledge, there is a lack of studies proposing, evaluating, or validating triggers for the detection of drug-induced respiratory disorders and related hospitalizations.7 Therefore, the aim of this pilot study was to propose and evaluate the performance of triggers in detecting drug-induced respiratory disorders related to hospitalizations in older people, as well as to estimate the prevalence of these ADEs in a Brazilian hospital.

Ethical aspects
This study was carried out by consulting electronic chart documentation with approval from the Research Ethics Committee of the São Paulo State University (UNESP) (CAAE 46938821.7.0000.5426).

Study design and setting
A six-month pilot cross-sectional study with retrospective data collection was conducted to propose and evaluate the performance of triggers for detecting hospitalizations related to drug-induced respiratory diseases.
The study was conducted at Am erico Brasiliense State Hospital (HEAB), located in the interior of the state of São Paulo, Brazil.The hospital has 84 beds and serves patients from the Regional Health Division III, which comprises 24 cities and approximately 989,971 inhabitants.
The reporting of this study was based on Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) for cross-sectional studies. 9

Participants
Older people (age ≥ 60) admitted to general and surgical wards of the hospital between January and June 2021 were eligible.These two medical wards correspond to 38 beds, where patients are admitted regardless of their age or specific health conditions, for the diagnosis or management of morbidities and infections and the performance of health procedures (e.g., medical exams and surgeries).
The primary exclusion criteria were older people without signs and/ or symptoms of respiratory disorders or those diagnosed with COVID-19, regardless of whether the diagnosis was made before or during hospitalization.Older people diagnosed with COVID-19 were not eligible due to the difficulty in distinguishing between signs and/or symptoms of the clinical condition and drug-induced respiratory disorders.
The secondary exclusion criteria consisted of older people who had not used drugs before hospital admission.
Hospitalizations and readmissions from the two wards were evaluated during the study period to obtain a robust sample.The flowchart of the participant recruitment process is shown in Fig. 1.

Screening for drug-induced respiratory disorders
The triggers were screened in the electronic chart documentation of the multidisciplinary team (physicians, pharmacists, nurses, and physical therapists) and in therapeutic prescriptions.
Given the absence of specific triggers for detecting hospitalizations related to drug-induced respiratory disorders, the following screening strategies were adopted: (i) International Code of Diseases 10 th Revision (ICD-10) codes: screening was carried out by identifying at least one of the codes referring to respiratory diseases, signs, or symptoms (Supplementary Material 1).
In view of the routine and time constraints of the medical team, 10 not all ICD-10 codes referring to diseases, signs, or symptoms were assigned in electronic chart documentation.To ensure the screening of these triggers, a researcher (G.S) assigned relevant ICD-10 codes for older people who had a diagnosis of respiratory diseases (e.g., asthma) or signs and/or symptoms of respiratory disorders (e.g., dyspnea).(ii) Triggers from Global Trigger Tool (GTT): 11 The potential triggers present in the GTT that suggest the presence of a potential druginduced respiratory disease are listed in Supplementary Material 2. (iii) Pharmacological classes and drugs associated with respiratory disorders: screening consisted of identifying drugs that are potentially associated with respiratory disorders according to scientific evidence. 12,13Their screening in the electronic chart documentation of older people with signs and/or symptoms of respiratory disorders suggests the presence of ADEs.Pharmacological classes and drugs used during the hospitalization were not considered as triggers (Supplementary Material 3).

Data collection and variables
Data collection was conducted by consulting the electronic chart documentation using a review form previously designed by the researchers (Supplementary Material 4).The following variables were recorded: Sociodemographic: Sex; age (young older people [age ≥60−79] and long-lived older people [age ≥ 80]; and self-declared color (white, brown ["Pardo" in Brazilian Portuguese], and black).

Drug-induced respiratory disorders assessment
Causality assessment of ADEs An ADE was defined as any harmful event occurring during the use of drug therapy, regardless of dose. 15A drug-induced respiratory disorder was defined as an ADE with signs and symptoms affecting the entire respiratory system (e.g., pulmonary parenchyma and airways). 16ince this study aimed to identify hospitalizations related to druginduced respiratory disorders, only drugs used before hospitalization were assessed.Additionally, during hospitalization, ADEs were managed, as clinical pharmacists and physicians conducted the medication review and assessed the risks versus benefits of pharmacotherapy.
Causality assessment was performed for all hospitalizations and readmissions using the instrument developed by the World Health Organization. 17The assessments were carried out by one researcher (Author 1) and supported by discussions with two other researchers (M.F and F. R. V).The choice of this instrument is justified by its greater consistency in imputing ADEs in hospitals. 18actors such as the temporal relationship between drug use and potential respiratory disorder, scientific evidence, pharmacological plausibility, and confounding variables (e.g., diagnosis of respiratory diseases and smoking habit) were considered.Causality assessments with the imputation as certain, probable, and possible were considered as potential drug-induced respiratory disorders.
Preventability and severity of ADEs This study assessed whether drug-induced respiratory disorders could have been avoided, based on the following definitions: 19 − Non-preventable: adverse drug reactions that cannot be prevented as they result from the intrinsic properties of the drug itself.− Preventable: medication errors that might lead to patient injury.
For preventable drug-induced respiratory disorders, severity was classified according to the medication error classification defined by the National Coordinating Council for Medication Error Reporting and Prevention. 20Only categories describing patient harm were considered: E (an error occurred that might have contributed to or resulted in temporary harm to the patient and required intervention) and F (an error occurred that might have contributed to or resulted in temporary harm to the patient and required initial or prolonged hospitalization).

Therapeutic ineffectiveness
The therapeutic ineffectiveness of certain drugs may induce or exacerbate respiratory signs, symptoms, and disorders.Ineffectiveness was defined as the absence or reduction of the expected therapeutic effectiveness of the drug under the prescribed or indicated conditions of use. 21Medical reports were evaluated to identify potential cases of therapeutic ineffectiveness associated with signs and/or symptoms of respiratory disorders.

Statistical methods
Sociodemographic, clinical, pharmacotherapeutic, and lifestyle variables were presented descriptively using absolute and relative frequency.For continuous variables, mean and standard deviation were reported.
To assess the statistical difference between the groups (older people with and without drug-induced respiratory disorder), Fisher's exact test for qualitative variables or the t-Student for quantitative variables was used.

Performance of the triggers
To evaluate the performance of the triggers in detecting druginduced respiratory disorders, the Positive Predictive Value (PPV) was calculated according to the following equation:

PPV
Number of older people with potential drug induced respiratory disorders detected by the triggers Number of older people with signs and or symptoms of drug induced respiratory disordersscreened by the triggers PPV was performed only for triggers that detected potential druginduced respiratory disorders, values ≥ 0.20 were considered a good performance. 22

Data accessibility
The supplementary material is available at Open Science Framework (doi:10.17605/OSF.IO/EY6MX).

Characteristics of older people
Between January and June 2021, 306 older people were admitted to the hospital in the study (346 hospitalizations).In the selected general and surgical wards, 221 older people were admitted (251 hospitalizations) and 72 older people were eligible (81 hospitalizations) (Fig. 2).The greater number of hospitalizations is explained by the fact that an older person may have been admitted to the selected wards more than once during the data collection period.
Among the 72 older people who met the eligibility criteria, most were male (n = 37), with a mean age of 72 years (SD ± 8), and a mean length of hospital stay of 11 days (SD ± 10).The most frequent diseases were high blood pressure (n = 51) and diabetes mellitus (n = 29).Thirty older people had a diagnosis of respiratory diseases, four had a history of previous respiratory infections, and three were diagnosed with respiratory infections during hospitalization.Most of the participants were abstainers (n = 49) and nonsmokers (n = 58) (Table 1).

Screening and performance of the proposed triggers
The 72 eligible older people had at least one trigger identified, with an average of six triggers screened (SD ± 1).The most screened triggers were the ICD-10 codes referring to respiratory diseases, signs, or symptoms (n = 214), mainly cough (n = 52) and dyspnea (n = 48), followed by pharmacological classes and drugs associated with respiratory disorders (n = 130).The most identified pharmacological classes during the screening were beta-blockers (n = 38) and calcium channel antagonists (n = 20), while the drugs were aspirin (n = 21) and hydrochlorothiazide (n = 11).The triggers from GTT were screened 18 times, especially the trigger "transfer to a higher level of care" (n = 14) (Supplementary Material 5).
The total number of screened triggers is greater than the number of hospitalizations, as a case of potential drug-induced respiratory disorder could be screened by more than one trigger (Fig. 2).

Causality assessment and prevalence of drug-induced respiratory disorder
A causality assessment was performed for 72 older people (81 hospitalizations), evaluating 319 drugs associated with potential respiratory disorders.
Causality assessment as probable or certain for drug-induced respiratory disorders was not possible for these six older people.Four older people had a diagnosis of respiratory diseases or a clinical condition justifying respiratory signs and/or symptoms, and two did not have reports of withdrawal and rechallenge of the drug associated with the potential respiratory disorder − criteria required to impute an ADE as probable or certain (Supplementary Material 6).
The assessment of the degree of causality between the occurrence of potential respiratory disorders and the use of pharmacotherapy is described in Supplementary Material 6.

Preventability and severity of ADEs
Potential drug-induced respiratory disorders could have been prevented in three older people, as the patients had a diagnosis of chronic obstructive pulmonary disease and had experience with the use of drugs associated with exacerbation of respiratory disorders before hospital admission (i.e., codeine, enalapril, atenolol, and clonazepam).This ADE was responsible for the hospital admission of two older people (category F), and in one case, the disorder was identified during the hospital admission for the investigation of neoplasm (category E).

Therapeutic ineffectiveness
None of the six older people with potential drug-induced respiratory disorders presented therapeutic ineffectiveness (6/72).Among the other 66 eligible older people (66/72), six had potential cases of therapeutic ineffectiveness, as reported by the physician.Four had pulmonary edema due to an inappropriate dosage of furosemide (underuse); one had dyspnea due to the use of lower doses of formoterol and budesonide; and one presented a cough due to the use of a lower dose of prednisone.
For the detection of therapeutic ineffectiveness induced by an inappropriate dosage of the drug, it was possible to propose two drugs as triggers: furosemide for detecting pulmonary edema (PPV = 0.23) and prednisone for detecting cough (PPV = 0.20) (Table 2).

Discussion
In this pilot study, the proposed triggers allowed the detection of drug-induced respiratory disorders related to hospitalizations in older people.The triggers "abrupt medication stop" and the "use of codeine, captopril, and carvedilol" demonstrated good performance and enabled the identification of medication errors with harm (categories E and F).Additionally, two drugs used as triggers showed good performance in detecting therapeutic ineffectiveness induced by inappropriate dosages (furosemide and prednisone use).
Although the ICD-10 codes were the most frequently screened triggers in electronic chart documentation and detected all cases of druginduced respiratory disorders, they demonstrate poor performance and a high rate of false positives.This might be explained by the ICD-10 codes referring mainly to the diagnosis of pre-existing respiratory diseases (e.g., asthma) and associated signs and/or symptoms (e.g., dyspnea).
One of the novelties of this study was the proposal to use pharmacological classes and drugs potentially associated with respiratory disorders as triggers. 12,13Considering their good performance, easy applicability, and the limited time available to health professionals in clinical practice, 23 these triggers might facilitate the detection and management of ADEs during the care process and reduce health problems.However, there are some caveats to be considered.While some triggers are more practical and specific for the detection of certain types of ADEs (e.g., naloxone for opioid intoxication), others are general and require additional time to detect ADEs (e.g., administration of antihistamines). 22,24ong the triggers from the GTT, "abrupt medication stop" demonstrated good performance in detecting dyspnea induced by atenolol.However, this trigger was detected only once, and its applicability in clinical practice may be limited since it requires a thorough review of all patient electronic chart documentation and potential causes of abrupt medication stops.Despite being widely used, 25 the triggers from the GTT demonstrated better performance in detecting other types of ADEs, mainly constipation and hypotension. 26Furthermore, it is not possible to compare the present findings with the literature, as to the best of our knowledge, there are no reports of the GTT's performance in detecting drug-induced respiratory disorders in older people.
The authors proposed two triggers for detecting therapeutic ineffectiveness.Underuse of furosemide was associated with exacerbation of heart failure resulting in pulmonary edema, while underuse of prednisone, budesonide, and formoterol was responsible for exacerbation of respiratory diseases. 27These triggers, like the pharmacological classes and drugs, are practical and specific and can be useful in clinical practice for detecting therapeutic ineffectiveness.
Hospitalized for eight days due to an exacerbation of dyspnea and cough.
Upon hospital admission, the use of azithromycin and codeine was discontinued, and the use of morphine and dexamethasone was initiated.An improvement in dyspnea was observed, which might be attributed to the suspension of codeine and the use of morphine and dexamethasone.
Comorbidities: chronic obstructive pulmonary disease, depressive disorder, diabetes mellitus, fibromyalgia, high blood pressure, and osteoporosis.History of exposure to a wood stove for 12 years.
Hospitalized for 10 days due for an investigation of lung neoplasm and chronic cough.
The enalapril dosage was reduced and the cough improves.
Possible for enalapril induced cough (category E).
Comorbidities: chronic obstructive pulmonary disease, diabetes mellitus, high blood pressure, and lung neoplasm.
Hospitalized for two days due an exacerbation of chronic obstructive pulmonary disease, dyspnea and cough.
Upon hospital admission, the use of clonazepam and atenolol was discontinued, and it was observed an improvement in respiratory symptoms.
Possible for atenolol and clonazepam induced dyspnea (category F).
Comorbidities: anxiety disorder, chronic obstructive pulmonary disease, and high blood pressure.
Hospitalized for 42 days due an exacerbation of cough and bronchopneumonia.
Upon hospital admission, captopril was discontinued, and amoxicillin was prescribed, with improvement in respiratory signs.
Hospitalized for 13 days due to an exacerbation of cough and dyspnea.
Upon hospital admission, carvedilol was discontinued and meropenem was prescribed, with improvement in respiratory signs.
Hospitalized for eight days for palliative care and management of bronchopneumonia and dyspnea.
Upon hospital admission, the use of clonazepam and trazodone was discontinued.The patient showed improvement in respiratory signs, but the patient evolves to death.
medication review associated with the use of triggers enables the optimization of drug therapy 23 and contributes to the design of a care plan for the prevention of iatrogenic cascade and the enhance of patient safety. 30evertheless, some aspects should be considered.Triggers need to be selected according to the prevalence of ADEs, performance (PPV), target population, health service, and available resources (e.g., data register in electronic chart documentation). 23In addition, the routine of health professionals may impair the screening of ADEs with triggers. 23The number of staff, time constraints, and non-automation of the process are relevant factors when selecting triggers to be incorporated into practice.
Moreover, the performance of triggers may be influenced by the prevalence of ADEs, the level of health service, differences in diagnostic, and therapeutic practices among health professionals, 31 and institutional protocols. 32These factors may explain the absence of a consensus in the literature about the PPV considered as the gold standard, with good performance PVV ranging from 0.05 to 0.50. 7In this study, PPV ≥ 0.20 was adopted as good performance due to the low prevalence of drug-induced respiratory disorders, their difficulty in detection, and their similarity with respiratory diseases.However, PPV ≥ 0.20 for other types of ADEs, such as falls and fractures induced by nervous system depressants, may be considered poor performance since these types of ADEs are prevalent in clinical practice. 33hrough the application of triggers, was estimating a prevalence of 8.3 % of older people admitted to a hospital with potential drug-induced respiratory disorders.This prevalence is similar to that reported in a study conducted by Woo and colleagues (2020), who identified a 7.5 % prevalence of drug-induced respiratory disorders among adverse drug reaction reports in a Korean database. 9The prevalence identified in the present study might be underestimated due to the small sample size and possible under-detection of drug-induced respiratory disorders, given the diagnosis of previous respiratory diseases in about half of the older people and missing data on the use of certain drugs influenced the imputation of ADEs.This study has strengths and limitations.As strengths, it is noteworthy that this study is a groundbreaking proposal, as there is no previous study that evaluated the performance of triggers for detecting hospitalizations related to drug-induced respiratory disorders.Additionally, this study strengthens pharmacovigilance signals, which are still a bottleneck in Brazil, since the detection of druginduced respiratory disorders was only possible through screening with the proposed triggers, as ADEs were not reported in electronic chart documentation.
As limitations, considering this was a pilot study, the sample size and analysis period should be highlighted.A total of 72 older people admitted to two wards in a Brazilian hospital were assessed over a six-month period and seasonality was not considered.Consequently, the findings may not reflect the reality of other hospitals or health facilities, limiting the inference of the data.To mitigate these possible biases, all hospitalizations during the study period were evaluated.In addition, another limitation was missing data, a barrier commonly found in studies that use secondary sources of information (electronic chart documentation). 7onsidering the aforementioned aspects, future research is needed to evaluate the performance and usefulness of triggers for detecting hospitalizations related to drug-induced respiratory disorders in older people.Furthermore, improving the reporting in electronic chart documentation is required, and applying these triggers with prospective data collection could be a viable strategy for detecting and preventing this type of ADE.

Conclusion
In this pilot study, one in 12 hospitalizations was related to cough and/or dyspnea induced by atenolol, captopril, carvedilol, clonazepam, codeine, enalapril, and trazodone.Half of the potential drug-induced respiratory diseases could have been avoided, as the older people had previous pulmonary impairment.Among the proposed triggers, "abrupt medication stop" and "the use of codeine, captopril, and carvedilol" showed good performance in detecting hospitalizations related to druginduced respiratory disorders, while furosemide and prednisone use were effective in detecting therapeutic ineffectiveness.
The results obtained in this study are promising, and further studies are required to assess the performance and usefulness of incorporating these triggers into clinical practice for screening, detection, management, and reporting of this type of ADE, which is generally underreported and difficult to detect.This could help reduce drug therapyrelated problems and enhance patient safety.

Fig. 1 .
Fig. 1.Screening of triggers and data collection and medication review in electronic chart documentation of older people admitted to two medical wards of Am erico Brasiliense Hospital, January to June 2021.ICD-10, International Code of Diseases-10 th Revision.

Fig. 2 .
Fig. 2. Flowchart of detection of potential drug-induced respiratory disorders in older people admitted to two medical wards of the Am erico Brasiliense Hospital, São Paulo, January to June 2021.ICD-10, International Code of Diseases-10 th Revision.

Table 1
Baseline characteristics of older people admitted to two medical wards of the Am erico Brasiliense Hospital (n = 72) from January to June 2021, Brazil.
95 % CI, Confidence Interval 95 %; OR, Odds Ratio; SD, Standard Deviation.a p-value obtained in the t-Student test.The p-value of other variables were obtained in the Fisher's exact test (the number of participants in one or more categories is less than five).

Table 3
Description of the clinical history of the six older people with a potential drug-induced respiratory disorders related to hospital admissions.