Surveillance of Antibiotic Consumption Using the “Focus of Infection” Approach in 2 Hospitals in Ujjain, India

Antibiotic surveillance initiatives are limited in resource-constrained settings. In the present study, a quantitative comparison of antibiotic use rates for suspected infections in 2 hospitals in India was performed using the “focus of infection” approach to identify targets for quality improvement in antibiotic prescription patterns in hospitalized patients. Methods This observational study was carried out in one teaching and one nonteaching hospital. All the patients with suspected bacterial etiology were included. Data on the prescribed antibiotics and the focus of infection were prospectively collected using a structured questionnaire. Each diagnosis was further reviewed and confirmed by an independent consultant. The prescribed antibiotics were coded according to the World Health Organization Anatomic Therapeutic Classification (ATC) index with the defined daily dose (DDD) methodology. Focus-specific DDDs were calculated per hundred patient days (DDD/HPD). Results A total of 6026 patients were included from 72 participating physicians out of available 75 physicians. Overall antibiotic prescribing was higher by 5 percentage points in the teaching hospital (95%) than in the nonteaching hospital (90%). Quinolones (ciprofloxacin constituting 86% of DDD/HPD) were the highest prescribed class in the teaching hospital, and third-generation cephalosporins (with ceftriaxone and ceftriaxone/sulbactam constituting 40% and 28% of the DDD/HPD, respectively), in the nonteaching hospital. The targets identified for improvement were the following: longer than recommended duration of prophylaxis and lack of distinction between prophylaxis and therapy among surgical patients; irrational antibiotic prescribing in gastroenteritis; overuse of quinolones and lack of use of penicillin in pneumonia; overuse of quinolones and lack of use of doxycycline and macrolides in genital infections; and overreliance on antibiotics for treating skin and soft tissue infections. Conclusions Providing a quantitative comparison of antibiotic use rates for suspected infections, using the “focus of infection” approach along with the ATC/DDD methodology, appears appropriate for identifying targets for quality improvement with regards to antibiotic prescribing.


Introduction
Antibiotic resistance is a rapidly increasing public health problem [1]. Although most of the evidence is at an ecological level, it is widely accepted that the most potent driver for antibiotic resistance is antibiotic use [2]. The availability and use of the World health Organization (WHO) Anatomic Therapeutic Clinical classification and defined daily dose (ATC/DDD) methodology facilitate meaningful comparisons of antibiotic consumption across hospitals and also between countries [3]. An increase in the use of antibiotics in hospital settings has been documented worldwide, with a simultaneous increase in resistance and spread of resistant strains of many bacteria [4][5][6][7].
Disease surveillance projects have been initiated in several countries, and national antibiotic policies have been formulated. These projects, which used the ATC/DDD methodology, include Strama in Sweden [8], Danish Integrated Antimicrobial Resistance Monitoring and Research Programme (DANMAP) in Denmark [9], Dutch Working Party on Antibiotic Policy (Dutch acronym is SWAB) [10], Surveillance of Antibiotic Use and Resistance in Intensive Care (SARI) or Medical Antibiotic Use Surveillance and Evaluation (MABUSE) in Germany [11], European Surveillance of Antimicrobial Consumption (ESAC) in Europe [4,12] and Intensive Care Antimicrobial Resistance Epidemiology (ICARE) in the United States [13]. However, there is a serious lack of similar initiatives in resource-constrained settings, where the burden of infections requiring effective antibiotics is higher [1,7]. Because the cost of health care for resistant infections is high, injudicious use of antibiotics is a greater public health problem with respect to quality of patient care in resource-constrained settings. The National Centre for Disease Control, under the Director General of Health Services, Ministry of Health and Family Welfare, Government of India, published The National Policy for Containment of Antimicrobial Resistance, India in 2011 [14]. However, there are surprisingly few published reports describing the use of antibiotics in hospitals in India using the WHO ATC/DDD methodology. This lack of information hinders discussion of targeted interventions to reduce irrational antibiotic prescribing.
Accurate evaluation of antibiotic use can be achieved using patient-level surveillance [15]. However, even in the presence of accurate computerized prescription data, linking prescribing information to a given patient or diagnosis might not be possible. Extracting information from handwritten case records is timeconsuming. In many resource-constrained settings, medical records are poorly maintained, leading to underestimation and misclassification of the underlying etiology associated with the prescription of antibiotics [16]. We have attempted to address these methodological challenges in the present study by using the ''focus of infection'' approach, along with the WHO ATC/DDD methodology, to study hospital antibiotic prescribing in a resourceconstrained setting. The aim of this study was to provide a quantitative comparison of antibiotic use rates for suspected infections, using the ''focus of infection'' approach to identify targets for quality improvement with regard to antibiotic prescribing, taking 2 hospitals in Ujjain, India, as examples.

Study settings
The study sites were 2 hospitals. One is a 570-bed teaching hospital attached to the RD Gardi Medical College, which is a nonpaying facility located approximately 6 kilometers from the city of Ujjain. The consultants in this hospital have an institutional hospital-based practice and are not allowed to work in private practice. The other hospital is a 330-bed nonteaching hospital located in the city of Ujjain. In this hospital, all services are charged, but the hospital is run on a break-even basis. The consultants in the nonteaching hospital are allowed to work in private clinics outside of their official hours of work. Both hospitals cater predominantly to a rural population from the villages surrounding Ujjain city. In both hospitals, most admissions (91%) in medical and intensive care units are emergency admissions, whereas in the surgical units, both elective and emergency admissions are equally common. Treatment guidelines for infectious diseases have not been implemented in either hospital.

Study participants
All the patients in whom the admitting consultant suspected an infectious etiology at the time of admission or during the hospital stay and for which antibiotic therapy was started were included in the study. The patients admitted for infectious etiologies requiring anti-infective agents other than antibiotics and those treated for tuberculosis were not included.
The participating departments or units in both hospitals were the departments of pediatrics, including the neonatal intensive care unit; general medicine, including the medicine intensive care unit; general surgery; obstetrics and gynecology; ear, nose, and throat; and orthopedics. A total 72 physicians out of available 75 from the above-mentioned departments participated in the study, a participation rate of 96%. Participation was voluntary.

Sample size
The sample size calculation was based on a pilot study done in March and May 2007, which estimated that between 80 and 90% of all the admitted patients were prescribed antibiotics. Assuming a prevalence of 80% for antibiotic prescribing and requesting 80% power to detect a difference of 10 percentage points between hospitals for a given focus of infection in a statistical test for the comparison of 2 proportions with a 5% significance level, the required sample size was 108 in each of the focus of infection to be compared. Since, we had 15 foci of infection to be compared the minimum sample size was 15 X 108 X 2 = 3240 patients.

Data collection form and procedure
The data collection form included patient details, age, sex, admission ward, dates of admission, discharge, prescribed antibiotic (s) (classified according to ATC), start and completion of antibiotic treatment course, dose per administration, number of doses per day, and route of administration. Any change in the antibiotic prescribed, its dose, or duration was noted. The final diagnosis was coded according to the targeted anatomical systems or subsystems identified as ''focuses of infection'' (Table 1) [4,17]. Examples of typical infections in each group are also listed. The same diagnosis codes were used for surgical prophylaxis and for therapy. The duration of therapy for surgical prophylaxis was noted as single dose, single day, or more than 1 day.
The questionnaire was attached to the inpatient files of all the admitted patients. The admitting consultant filled in the diagnosis codes. Each focus of infection was categorized into 1 of 3 indications for therapy or prophylaxis: (a) community-acquired infection; (b) hospital-acquired infection (for definition, see table 1); or (c) perioperative or medical prophylaxis.
The resident medical officers of the participating departments filled in the details of antibiotics prescribed by following up each patient from admission until discharge. A second independent consultant reviewed all the forms and also discussed and resolved any controversies regarding the focus of infection. The clinical suspicion of a focus of infection was confirmed by appropriate hematological, biochemical, and radiological investigations.
The structured questionnaire used in the present study was adapted from the Strama point prevalence surveys and ESAC [4,17]. The final form was pilot tested twice in March and May 2007 on 100 patients each in both hospitals.
Each patient admission was counted only once, and patient transfers between units of a hospital were not counted as separate admissions. A patient's duration of stay was calculated by subtracting the date of admission from the date of discharge. The days of admission and discharge were counted together as 1 day. The same data collection form and procedure were used in both hospitals. Only 1 focus of infection, which was considered as the most relevant for therapy or prophylaxis, was included.

Duration of data collection
Data collection was done from November 2007 to February 2009. The data were collected for 45 days in each hospital. A gap of 15 days was introduced between the 2 hospitals and every 4 months to allow time for appropriate data management. The study period covered 4 seasons, namely, 2 winters (1 in the beginning and 1 at the end, November to February), 1 summer (March to June), and the rainy season (July to October).

Data management and statistical analysis
Each prescribed antibiotic was coded according to the WHO Collaborating Centre for Drug Statistics Methodology, ATC classification index with DDDs 2009 [3] as defined in the ATC fifth level in J01 (antibacterial for systemic use). Nitroimidazole derivatives (P01AB) and nitazoxanide (P01AX11) were also included. Focus-specific DDDs were calculated per hundred patient days (DDD/HPD). The DDD/HPD of a given focus of infection was compared to study the antibiotic prescribing pattern of the 2 hospitals.
The data were entered into the EpiData Entry (version 3.1) and then transferred to the Stata 10.0 for further analysis (Stata Corp., College Station, Texas, USA). A descriptive data analysis was conducted to understand the frequency of the patient-related variables (age groups, sex, services of the hospital, and season), foci of infection, and prescribed antibiotic and its class. The Pearson chi-square test was used to test for statistical significance (5%).

Ethics statement
The ethics committee of RD Gardi Medical College approved the study (approval no. 41/2007). Informed written consent was obtained from all the patients. The research was conducted according to the principles expressed in the Declaration of Helsinki. The study did not interfere with the ongoing treatment of the patients.

Results
The study included 6026 admitted patients, 2525 (42%) of whom were women and 3501 (68%) were men; 2352 patients (39%) were from the teaching hospital, and 3674 (61%) were from the nonteaching hospital. A total of 5531 patients (92%) were prescribed antibiotics for various foci of infection. Table 2 shows the details of the foci of infection recorded for the antibiotic treatment for the community-acquired infections, hospital-associ-ated infections (HAI), and prophylaxis. Prophylaxis was given to 1846 patients (34% of the total), and in 86% (n = 1593), it was given for more than 1 day. Eighty-six percent of those on prophylaxis for more than 1 day belonged to surgical services, including the departments of general surgery; obstetrics and gynecology; ear, nose, and throat; and orthopedics. Gastroenteritis, pneumonia, bronchitis, central nervous system, and cardiovascular system infections and undifferentiated febrile illnesses were predominantly community-acquired. HAI was identified in 9% of all the infections. The most common HAI recorded was skin and soft tissue infection due to surgical site infection.
The patient's age group, sex, the hospital service, and the season in which the patient was admitted were significantly associated with antibiotic prescribing (Table 3). Most patients (59%) were prescribed 1 antibiotic. A combination of 2, 3, or 4 antibiotics was prescribed significantly more often in the teaching hospital. The mean length of stay in the teaching hospital was 7.8 days (95% confidence interval [CI] 7.3-8.3 days) and that in the nonteaching hospital was 4.2 days (95% CI 4.1-4.4 days).
The prescribing pattern for bronchitis was similar to that for pneumonia in the respective hospitals (Table 4 and Figure 1).
Renal infections. In the treatment of renal infections, the antibiotic prescribing pattern was similar in the 2 hospitals. The quinolones, third-generation cephalosporins, and aminoglycosides were the most common classes prescribed. In the nonteaching hospital, however, a greater proportion of levofloxacin, ceftriaxone with a beta-lactam inhibitor, amikacin, and nitelmycin were prescribed (Table 4).

Discussion
To the best of our knowledge, this is the first study to use the ''focus of infection'' approach, along with the WHO ATC/DDD methodology, for quantifying antibiotic prescribing in India. The DDDs were calculated per 100 patients per focus of infection per day (DDD/HPD). The results show that more DDD/HPD were prescribed in the teaching hospital for nearly all the foci of infections. These findings are similar to those of a large French study that demonstrated that higher antibiotic DDDs were dispensed in teaching hospitals [5]. The higher rate of antibiotic use in the teaching hospital in our study could be due to its status as a referral center for complicated cases, although disease severity was not monitored in our study. Newly marketed antibiotics like the combination of third-generation cephalosporins with betalactamases and Nitazoxanide, were prescribed more often in the non-teaching hospital. Since, in the non-teaching hospital services are provided for a cost and the teaching hospital is a free-of-cost facility, the payment status of the hospital is likely to influence the antibiotic prescribing. In a countrywide study among Irish general practitioners (GPs) it was found that GP's decision to provide a prescription for antibiotics might be influenced by whether or not the patient pays for their consultation {Murphy, 2011}. Combination antibiotic therapy was more common in the teaching hospital. However, the combination of third-generation cephalosporins with beta-lactamases, which is not a rational choice, was used more often in the nonteaching hospital. There is evidence that monotherapy is sufficient for all serious infections without shock [18]. Thus, developing guidelines for prudent antibiotic use for common infections is a priority identified through this study.
Approximately 9% of the infections were classified as HAI, which is a lower incidence than the expected rate of around 25% [19]. This might be due to the lack of specific surveillance for identifying HAIs and also to the lack of awareness of HAI. SSIs were the most commonly detected HAIs, and 28% of all the SSIs were surgical site infections. SSI was also a common reason for surgical prophylaxis. Establishing surveillance for HAI and specifically for SSI was also identified as a target for quality improvement in this study.
Approximately one third of all antibiotic prescribing was for prophylaxis, mainly in surgical services. Most (86%) of these patients received antibiotics for longer than 24 hours, which is a cause of concern that should be considered for possible intervention. Unnecessarily long surgical prophylaxis was also reported in the ESAC hospital point prevalence studies in 2006, 2008, and 2009 [12].
Variation in antibiotic prescribing according to age groups was observed in this study, with higher rates (by 4 to 9% unit points) of antibiotic prescribing in adults than in children. Such variation     was documented previously in the same setting in an outpatient diagnosis prescribing study [20]. We analyzed the pattern of antibiotic prescribing in specific foci of infection. Over prescription of antibiotics for diarrhea is a major public health problem in India [21]. Acute diarrhea is self-limiting in most cases, and treatment should be restricted to rehydration, correction of electrolyte imbalance, and oral zinc for children [21]. Prescription of the nitroimidazole group of antibiotics, especially in children, cannot be explained rationally. The reasons behind such prescribing need to be researched, and appropriate, contextspecific interventions need to be introduced.
Prophylactic use of antibiotics is common in patients undergoing surgical procedures [22]. A total of 508 (60%) of 844 patients received prophylactic antibiotics for prevention of infection localized to the upper and lower gastrointestinal tract. The most common antibiotic regimens were ciprofloxacin with metronidazole and third-generation cephalosporins with metronidazole in the teaching and nonteaching hospitals, respectively. Given the polymicrobial nature of intra-abdominal infections arising from resident enteric flora, the above-mentioned combination therapy might be justified but only as a treatment option for source control [22]. We observed no difference in the antibiotics chosen for treatment and prophylaxis. In addition, as already discussed, most prophylaxis continued beyond 24 hours after surgery. The lack of distinction between the use of antibiotic therapy for prophylaxis and its use for treatment of an established infection is identified as an important quality improvement target.
Doxycycline and tetracycline dominated the prescription pattern for pneumonia in the teaching hospital, and these are appropriate for community-acquired pneumonia (CAP) of all severities according to the British Thoracic Society guidelines [23]. The use of ciprofloxacin, which was the second most common antibiotic prescribed for CAP, is not appropriate especially in view of its poor activity against Streptococcus pneumoniae and Klebsiella pneumoniae, the 2 organisms most frequently associated with adult pneumonia in the Asia-Pacific region [24]. Non-use of penicillin is a cause of concern, as recent evidence shows that fewer than 5% of all the nonmeningeal isolates of S. pneumoniae from the Asia-Pacific are penicillin resistant (according to an MIC of 8 mg/mL) [24]. Convincing physicians to increase use of penicillin is identified as a key quality improvement issue.
The Infectious Disease Society of America recommends the use of 1 of the following 3 initial therapies for acute pyelonephritis in adults: (a) fluroquinolone, (b) aminoglycoside with or without ampicillin, or (c) extended spectrum cephalosporin with or without aminoglycoside [25]. The antibiotic prescribing pattern in renal infections appeared to be appropriate in both hospitals.
There is a significant unmet demand for the treatment of genital infections among women in rural India due to their high disease burden [26]. In addition, the skewed urban distribution of healthcare workers contributes to this unmet demand [27]. The optimization of therapy for hospitalized genital infections (especially pelvic inflammatory disease) must take into account the polymicrobial etiology of the disease, the severity of the disease, and patient compliance with antibiotic use. The optimal treatment for genital infections should include an antibiotic with activity against Neisseria gonorrhoeae, Chlamydia trachomatis, and Mycoplasma genitalium. Because no single agent covers all of these organisms, combination therapy is recommended [28]. In the teaching hospital, ciprofloxacin or norfloxacin were prescribed, often in combination with metronidazole. However, ciprofloxacin is less effective against bacterial vaginosis-associated microorganisms; therefore, possible relapse of the infection is of concern. In addition, increasing quinolone resistance among N. gonorrhoeae isolates has been documented in the Asia-Pacific region [7]. Thirdgeneration cephalosporins are a good addition to the regimen [28], which was used more often in the nonteaching hospital. Overuse of aminoglycosides and lack of use of doxycycline and macrolides were identified as quality improvement targets in antibiotic prescribing for genital infections.
Among hospital patients with SSI, antibiotic therapy should be initiated only if the patient fails to respond to incision and drainage or shows abscess with severe and extensive disease, rapidly progressive cellulitis, signs and symptoms of systemic disease, associated comorbidities, immunosuppression, or an abscess in an area difficult to drain (face, neck, or hand) [29]. Most of the patients treated with antibiotics in the 2 hospitals did not fulfill these criteria. The most common choice of antibiotic was cotrimoxazole, which appears appropriate. However, the relatively common use of quinolones and third-generation cephalosporins is not justified [29].
The main strength of the present study is that unlike most surveillance [4,5,6,11,12], which collects and presents dispensing data at aggregate levels, we collected information on individual patients and on antibiotics actually administered to the patients. Therefore, our data are robust and relevant for interventions and for monitoring trends. The validity of our data is strong because an independent expert confirmed the foci of infection. However, there are a few limitations. The design of the study is resource intensive, requiring dedicated personnel. We did not evaluate the severity of illness, which would have further intensified the workload. The study did not include positive bacterial cultures or other laboratory measures to confirm the focus of infection, but would like to do so in our future studies. Monitoring of sensitivity patterns itself can change prescription patterns. The results are not strictly comparable within a season, as data were not collected simultaneously in the 2 hospitals. In the present study we have focused on choice of antibiotics as the main outcome but other aspects of rational antibiotic prescribing like duration of treatment and appropriateness of choice of formulations (for example oral therapy versus intravenous and use of syrup or tablets) is not discussed in the study. The expectancy effect i.e. the consultants under observation changing the rate of antibiotic prescribing, is a potential bias. However, the prescribing was observed for a longer duration in this study, thus minimizing the expectancy effect.

Conclusions
Using a data collection procedure that can produce good focus of infection-specific information on ATC/DDD, we were able to identify targets for quality improvement in antibiotic prescribing. The targets identified are higher antibiotic prescribing in a teaching hospital compared with a nonteaching hospital, longer than recommended duration of prophylaxis and lack of distinction between prophylaxis and therapy among surgical patients, irrational antibiotic prescribing in gastroenteritis, overuse of quinolones and lack of use of penicillin in pneumonia, overuse of quinolones and lack of use of doxycycline and macrolides in genital infections, and overreliance on antibiotics in treating skin and soft tissue infections. The study provides much needed Indian data for policy makers to design strategies for promoting prudent antibiotic use and formulating national, regional, and local therapeutic guidelines.