Antibiotic‐associated suspected adverse drug reactions among hospitalized patients in Uganda: a prospective cohort study

Abstract We sought to determine the prevalence at admission and incidence during hospitalization of antibiotic‐associated suspected adverse drug reactions (aa‐ADRs) among Ugandan inpatients; and to characterize these aa‐ADRs. We conducted a prospective cohort study of 762 consented adults admitted on medical and gynecological wards of the 1790‐bed Mulago National Referral Hospital. Thirty percent were known HIV‐seropositive (232/762). Nineteen percent (148/762; 95% CI: 17–22%) of inpatients experienced at least one aa‐ADR. At hospital admission, 6% (45/762; 95% CI: 4–8%) of patients had at least one aa‐ADR; and 15% (45/300; 11–20%) of those who had received antibiotics in the 4‐weeks preadmission. Twenty‐four (53%) of these 45 patients had serious aa‐ADRs. The incidence of aa‐ADRs was 19% (117/629; 95% CI: 16–22%) of patients who received antibiotics [community‐acquired: 9% (27/300; 95% CI: 6–13%); hospital‐acquired: 16% (94/603; 95% CI: 13–19%)]: 39 (33%) of 117 patients had serious aa‐ADRs. Of 269 aa‐ADRs, 115 (43%) were community‐acquired, 66 (25%) probable/definite, 171 (64%) preventable, 86 (32%) serious, and 24 (9%) rare. Ceftriaxone was the most frequently implicated for serious hospital‐acquired aa‐ADRs. Cotrimoxazole, isoniazid, rifampicin, ethambutol, and pyrazinamide were the most frequently linked to serious community‐acquired aa‐ADRs. Fatal jaundice (isoniazid), life‐threatening difficulty in breathing with shortness of breath (rifampicin) and disabling itchy skin rash with numbness of lower swollen legs (ethambutol, isoniazid) were observed. Pharmaceutical quality testing of implicated antibiotics could be worthwhile. Periodic on‐ward collection and analysis of antibiotic‐safety‐data standardized by consumption is an efficient method of tracking antibiotics with 1%‐risk for serious aa‐ADRs.

Developing countries contributed 76% of the global rise in antibiotic use between 2000 and 2010 (Van Boeckel et al. 2014), which increases their risk of aa-ADRs. Cotrimoxazole use is standard-of-care for prophylaxis against opportunistic infections among HIV/AIDS patients in resource-limited settings (WHO, 2006a). Thus, the high burden of HIV/AIDS in sub-Saharan Africa (SSA) implies increased risk of cotrimoxazole-linked ADRs (Mouton et al. 2015). Little is known about the frequency (Kiguba et al. 2015) and characteristics of aa-ADRs among inpatients in SSA (Mehta et al. 2008), particularly in Uganda. Moreover, it is a challenge to ascertain aa-ADR causality in our hospital setting where a large proportion of inpatients have comorbidities and/or concurrently receive multiple medicines.
Four-fifths of inpatients at a tertiary care public health facility in Uganda consume at least one antibiotic during hospitalization, whereas two-fifths use at least one antibiotic during the 4-weeks preadmission (Kiguba et al. 2016a). Patients can receive parenteral antibiotics at private clinics in the community prior to their subsequent referral to tertiary care public health facilities.
The reported antibiotic prescribing patterns in our hospital setting excluded antituberculous drugs and focused on commonly used antibacterial agents (Kiguba et al. 2016a), whose prescription is not restricted to specialists, to identify key areas for antibiotic stewardship. For the benefit of antituberculous drugs pharmacovigilance in Uganda, this paper includes data on the aa-ADRs of antituberculous drugs. We sought to determine the prevalence at hospital admission and the incidence during hospital stay of aa-ADRs among Ugandan inpatients: their seriousness, rarity, preventability, causality, and severity.
Patients gave written informed consent.

Data collection
From October to November 2013, we undertook a pilot study on the four wards to assess practicability and to pretest the study instruments. The main study (reported here) was implemented in December 2013 to April 2014. Four trained ward-teams, each having a medical doctor, pharmacist and degree-nurse, were to recruit and followup inpatients using a systematic random sampling procedure: three new admissions daily on long-stay wards (HNE/CPN) and six on short-stay wards (IDGI/GYN). Ward-teams randomly selected the first study patient from the first two (IDGI), three (HNE), and four (CPN/ GYN) new admissions; and subsequently aimed at recruiting every second, third, and fourth admission, respectively (Kiguba et al. 2016a). Baseline patient assessment captured relevant data on demographics, clinical conditions including aa-ADRs, and antibiotic medications used. Subsequently, daily assessments were conducted until discharge, transfer, death, or loss to follow-up. Ward-teams collected data from 8.00am to 6.00 pm from Monday to Friday and from10.00am to 6.00 pm on weekends and public holidays (Kiguba et al. 2016a).

Data management
The data were double-entered into a database using Epidata 3.1 software (Odense, Denmark) with check programs to limit out-of-range data entry errors. Where data discrepancies occurred, the original case report form was cross-checked and corrections were made.

Identification of aa-ADRs
We defined aa-ADRs according to the WHO definition (WHO-UMC, 2011b). Clinical examination was the major approach used to identify aa-ADRs due to limitations in timely availability of laboratory investigation results (Kiguba et al. 2016b). To increase the probability to detect aa-ADRs, patients were screened using an ADR trigger tool (Rozich et al. 2003). To assess causality, a suspected aa-ADR was assigned to a Naranjo ADR probability category based on a total score obtained from 10 weighted questions. These questions assessed the temporal association between suspected drug and adverse reaction, alternative cause(s) of the reaction, plasma drug levels (if available), dose-response relationships and previous patient experience with the drug. Suspected aa-ADRs with Naranjo score of 0 were doubtful, 1-4 possible, 5-8 probable, and ≥9 definite (Naranjo et al. 1981). Thus, coding an adverse event as "aa-ADR" required at least possible grading on the Naranjo scale. Operationally, an aa-ADR was any undesirable medical occurrence that developed after the administration of an antibiotic and for which there was, at least, possible causality between the antibiotic and the medical occurrence. Consensus agreement on aa-ADR causality was reached in a committee headed by the ward-based study physician and senior clinical pharmacist (RK). This team approach reflected the routine on-ward approach whereby nurses, medical doctors, and clinical pharmacists brainstorm on patients' clinical problems before making clinical decisions (Kiguba et al. 2016a). Community-acquired aa-ADRs were defined as ADRs linked to preadmission use of antibiotics. Some community-acquired aa-ADRs manifested preadmission, whereas others occurred after hospital admission. Hospitalacquired aa-ADRs were those linked to hospital-initiated antibiotics used during the current hospitalization.
Preventability, severity (grade or intensity), and seriousness (incapacitating or life-threatening) of aa-ADRs were also determined by consensus as described above. Preventability was assessed using the modified Schumock and Thornton Preventability Scale (Schumock and Thornton 1992;Lau et al. 2003), whereas severity was evaluated using the Division of AIDS Table for Grading the Severity of Adult and Paediatric Adverse Events (Division of AIDS (DAIDS), 2004) and seriousness using the WHO Uppsala Monitoring Centre (UMC) criteria (WHO-UMC, 2000). Rarity of an aa-ADR (occurrence in <0.1% of medication users) (WHO-UMC, 2011b), was assessed by RK using the British National Formulary (BNF) (British National Formulary, 2014) as the principal reference.

Statistical analysis
We determined the prevalence of aa-ADRs at hospital admission, and the incidence of aa-ADRs during hospitalization. Numerators for prevalence and incidence were the number of patients who had experienced preadmission aa-ADRs and new cases of in-hospital aa-ADRs, respectively, whereas the denominator was the number of study patients who received antibiotics (both incidence and prevalence) or total number of patients in the cohort (prevalence only).
We also computed the incidence of hospital-acquired aa-ADRs per 100 defined daily doses (DDDs) (Anatomical Therapeutic Chemical/Defined Daily Dose (ATC/DDD) Index, 2015; Hamad et al. 2013) of each implicated antibiotic administered during the current hospitalization.
We assessed aa-ADRs as community-acquired or hospital-acquired; and for causality, preventability, severity, seriousness, and rarity.

Ethical clearance
Ethical approval for the study was obtained from the School of Medicine Research and Ethics Committee, Makerere University College of Health Sciences (REC REF No. 2011-113), the Mulago Hospital Research and Ethics Committee (MREC 253), and the Uganda National Council for Science and Technology (HS 1151).
A quarter of the patients who were admitted with at least one ADR related to any medication class had  prevalent community-acquired aa-ADRs. (15/62; 24%, 95% CI: 14-37%).

Rarity of aa-ADRs
Nine percent (24/269) of aa-ADRs, see Tables 7, S4, experienced by 22 patients were rare. Twelve rare aa-ADRs were attributed to metronidazole, those in italics being of the central nervous system [dizziness (4), constipation (3), convulsions (1), blurred vision (1), facial itchy skin rash (1), fever (1), and headache (1)]; and six to cotrimoxazole [anorexia (3), jaundice (1), dry cough with shortness of breath (1), and anemia (1)]. Nine of the 24 rare aa-ADRs were serious: four linked to cotrimoxazole (jaundice, anorexia, anemia, and dry cough with shortness of breath), three to metronidazole (blurred vision, convulsions, and dizziness), one to levofloxacin (peripheral neuropathy), and one to ceftriaxone (decreased urine output). Moderate jaundice (cotrimoxazole) and severe decreased urine output (ceftriaxone) were nonpreventable rare and serious aa-ADRs of probable causality, whereas the rest of the rare and serious aa-ADRs were of possible causality, see Table S4. The jaundice linked to cotrimoxazole occurred in a 33-year-old HIV-positive female who had not yet commenced antiretroviral therapy; the patient's other working diagnoses were acute gastroenteritis and dysentery. The severe decreased urine output linked to ceftriaxone occurred in an 18-year-old HIV-positive male not yet on antiretroviral therapy whose other working diagnosis was bronchopneumonia: This patient's final diagnosis was acute kidney injury.
Half the rare aa-ADRs (12/24) were communityacquired, eight of which were cotrimoxazole-linked and four metronidazole-linked. Eight of the 12 hospitalacquired rare aa-ADRs were attributed to metronidazole use and three to ceftriaxone, see Table S4.

Discussion
Suspected aa-ADRs were common both at admission and during hospital stay. The incidence of aa-ADRs in our study was threefold higher than that observed in a French cohort of 3963 hospitalized patients (Courjon et al. 2013) which could, in part, be linked to the higher consumption of antibiotics by our inpatients for HIV-associated comorbidities. Most of our inpatients received antibiotics whose use was significantly higher among HIV-infected patients (Kiguba et al. 2016a). The medical wards showed a higher HIV-seroprevalence (38%) than gynecology where HIV-seroprevalence (9%) was similar to the national estimate (7.3%) (Uganda AIDS Commission, 2015). The fact that most patients received antibiotics makes our study epidemiologically efficient. Few aa-ADRs would be observed in a hospitalized-population in which antibiotic prescribing was infrequent. Of all suspected ADRs attributed to any medication class, a large proportion (one in three) was linked to antibiotics, which corroborates findings from South Africa, a setting with a similarly high HIV/AIDS burden, where one in five ADRs (11/51) was antibiotic-associated (Mehta et al. 2008).
Ceftriaxone showed the strongest signal for hospitalacquired aa-ADRs, even among HIV-negative/unknown serostatus patients, which could be a reflection of ceftriaxone's high prescription rate in our hospital setting. Strong signals were also shown by metronidazole and levofloxacin (based on small numbers of patients for levofloxacin). Unmasking of the signal from the infrequently used levofloxacin demonstrates the utility of standardizing absolute aa-ADRs by DDDs administered (Hamad et al. 2013). However, much larger well-designed epidemiological studies in Uganda and other similar resource-limited settings could provide more clues on the aa-ADR profiles of ceftriaxone and levofloxacin, which are otherwise generally known to be safe. We did not test the quality of the antibiotics used and hence how this impacted on the incidence of hospital-acquired aa-ADRs by individual antibiotics. Nonetheless, this study has identified individual antibiotics (ceftriaxone, metronidazole, levofloxacin) with potentially higher risk for hospital-acquired aa-ADRs. The   with multiple comorbidities and on multiple medications. Hence, we screened patients using an ADR trigger tool to increase the probability to detect aa-ADRs; and subsequently applied the Naranjo ADR scale to the suspect clinical signs and symptoms to identify those that were at least possibly antibiotic-associated. We held extensive ward-team discussions and made consultations with the literature before reaching consensus on ADR causality. Sixth, we cannot claim generalizability of our findings to district hospitals, clinics, and other lower level health centers in Uganda. Similar studies conducted in lower level health facilities could elucidate the extent and characteristics of aa-ADRs in those settings.

Conclusion
Ceftriaxone showed the highest risk for hospital-acquired aa-ADRs (also of serious hospital-acquired aa-ADRs), even in HIV-negative/unknown serostatus patients, probably due to its high consumption. Levofloxacin signaled as high-risk for hospital-acquired aa-ADRs despite its low frequency of in-hospital use, thus, highlighting the utility of standardizing absolute aa-ADR counts by DDDs administered. Cotrimoxazole and antituberculous drugs were the most frequently implicated in communityacquired aa-ADRs (also in serious community-acquired aa-ADRs). Pharmaceutical quality testing of implicated antibiotics in Uganda could be worthwhile. In sentinel settings, periodic on-ward collection and analysis of antibiotic-safety-data standardized by DDDs, such as during a 6-week surveillance period once every 6 months, should be an efficient method of tracking antibiotics with 1%-risk for serious aa-ADRs.

Author Contribution
RK conceived of the study and drafted the manuscript and, in conjunction with SMB, participated in its design, implementation, statistical analysis, and drawing of inferences. CK participated in study design and, together with SMB, took part in the manuscript writing process. All authors approved the final manuscript.