Articles Use of standardised patients to assess antibiotic dispensing for tuberculosis by pharmacies in urban India: a cross-sectional study Srinath Satyanarayana, Ada Kwan, Benjamin Daniels, Ramnath Subbaraman, Andrew McDowell, Sofi Bergkvist, Ranendra K Das, Veena Das, Jishnu Das*, Madhukar Pai* Summary Background India’s total antibiotic use is the highest of any country. Patients often receive prescription-only drugs Lancet Infect Dis 2016; directly from pharmacies. Here we aimed to assess the medical advice and drug dispensing practices of pharmacies 16: 1261–68 for standardised patients with presumed and confirmed tuberculosis in India. Published Online August 24, 2016 http://dx.doi.org/10.1016/ Methods In this cross-sectional study in the three Indian cities Delhi, Mumbai, and Patna, we developed two S1473-3099(16)30215-8 standardised patient cases: first, a patient presenting with 2–3 weeks of pulmonary tuberculosis symptoms (Case 1); See Comment page 1208 and second, a patient with microbiologically confirmed pulmonary tuberculosis (Case 2). Standardised patients were *Contributed equally scheduled to present each case once to sampled pharmacies. We defined ideal management for both cases a priori as McGill International referral to a health-care provider without dispensing antibiotics or steroids or both. Tuberculosis Centre & Department of Epidemiology, Findings Between April 1, 2014, and Nov 29, 2015, we sampled 622 pharmacies in Delhi, Mumbai, and Patna. Biostatistics and Occupational Health, McGill University, Standardised patients completed 1200 (96%) of 1244 interactions. We recorded ideal management (defined as referrals Montreal, QC, Canada without the use of antibiotics or steroids) in 80 (13%) of 599 Case 1 interactions (95% CI 11–16) and 372 (62%) of (S Satyanarayana MD, 601 Case 2 interactions (95% CI 58–66). Antibiotic use was significantly lower in Case 2 interactions (98 [16%] of 601, A McDowell PhD, M Pai MD); 95% CI 13–19) than in Case 1 (221 [37%] of 599, 95% CI 33–41). First-line anti-tuberculosis drugs were not dispensed Manipal McGill Center for Infectious Diseases, Manipal in any city. The differences in antibiotic or steroid use and number of medicines dispensed between Case 1 and Case 2 University, Manipal, India were almost entirely attributable to the difference in referral behaviour. (M Pai); Development Research Group, The World Bank, Interpretation Only some urban Indian pharmacies correctly managed patients with presumed tuberculosis, but most Washington, DC, USA (A Kwan MHS, B Daniels MS, correctly managed a case of confirmed tuberculosis. No pharmacy dispensed anti-tuberculosis drugs for either case. Absence J Das PhD); Division of of a confirmed diagnosis is a key driver of antibiotic misuse and could inform antimicrobial stewardship interventions. Infectious Diseases, Brigham and Women’s Hospital and Harvard Medical School, Funding Grand Challenges Canada, Bill & Melinda Gates Foundation, Knowledge for Change Program, and World Boston, MA, USA Bank Development Research Group. (R Subbaraman MD); ACCESS Health International, Copyright © The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY license Hyderabad, India (S Bergkvist MS); Institute for Social and Economic Research Introduction common, non-specific, non-severe, and persistent. In on Development and Antimicrobial resistance is a global health emergency,1,2 this case, assessment of pharmacist behaviour provides a Democracy, Delhi, India and the indiscriminate use of antibiotics is a major realistic and externally valid estimate of unnecessary (R K Das PhD); Department of Anthropology, Johns Hopkins driver.3,4 Although India ranks first in total antibiotic use antibiotic use. Further, indiscriminate drug use can harm University, Baltimore, MD, USA worldwide,2 the absence of data linking antibiotic use to both the patient and the efficacy of existing anti-tuber- (V Das PhD); and Center for underlying illnesses makes it hard to assess the culosis treatments. For instance, tuberculosis symptoms Policy Research, New Delhi, appropriateness of such use in view of India’s high subside temporarily with the use of fluoroquinolones or India (J Das) infectious disease burden. With some of the highest corticosteroids, delaying diagnosis and leading to the Correspondence to: Prof Madhukar Pai, incidences of drug-resistant bacterial pathogens in the possibility that patients receive several antibiotic courses Canada Research Chair in world, identification of the sources and circumstances of for the wrong diagnosis.5 Partial courses of anti- Epidemiology & Global Health, antibiotic abuse as opposed to use is a crucial first step to tuberculosis drugs can result in drug resistance.6 Finally, McGill University Department of understanding what can be done about antibiotic overuse international and national guidelines for the optimum Epidemiology & Biostatistics, Montreal, QC H3A 1A2, Canada in India.4 Here, we develop a unique method to address management of tuberculosis cases7,8 allow the assessment madhukar.pai@mcgill.ca this gap, focusing our attention on a specific illness, of the extent of antibiotic misuse. tuberculosis, and a specific source of health care— Our focus on pharmacies is premised on the belief that pharmacies. their practices contribute to the availability and use of Our choice of tuberculosis, a disease that affects antibiotics in the population.4 This premise is partly the 2·2 million Indians every year, as a lens through which result of their widespread availability—more than to investigate antibiotic use is driven by several factors. 750 000 private retail pharmacies provide easy access to The symptoms of early pulmonary tuberculosis are drugs.9 However, our premise also reflects the willingness www.thelancet.com/infection Vol 16 November 2016 1261 Articles Research in context Evidence before this study symptoms (Case 1); and second, a patient with Antimicrobial resistance is a global health emergency, and, as microbiologically confirmed pulmonary tuberculosis (Case 2). the largest consumer of antibiotics, India is at highest risk. The Across all interactions, 319 (27%) of 1200 (95% CI 24–29) standardised patient method can help to assess the extent and resulted in the use of an antibiotic although no pharmacy appropriateness of antibiotic use because such use can be dispensed first-line anti-tuberculosis drugs. Ideal case directly related to the underlying illness of the patient. To management, defined as referrals without the use of antibiotics identify previous research on this topic, we searched PubMed or steroids, was much lower in Case 1 interactions (13%) than and Google Scholar using a combination of the terms Case 2 interactions (62%). Our study results add to the growing “standardized patients” (“mystery clients”, “fake patients”, or evidence on antibiotic abuse, but also underscore that the use “simulated patients”), “pharmacy” (“pharmacist” or “chemist”), and misuse of antibiotics are mediated by drug category and and “tuberculosis” with and without the keyword “India” for the information that patients present. Although antibiotic use articles published in English until March 31, 2015. Our search is high and such use can delay diagnosis, none of the showed that previous studies of physician management of pharmacies dispensed first-line anti-tuberculosis drugs, and the standardised patients in India have reported unnecessary use of stronger fluoroquinolone antibiotics and heavily antibiotic prescribing for various conditions, including restricted drug classes was low. Furthermore, the use of all tuberculosis, diarrhoea, asthma, and angina. However, these antibiotics decreased sharply when the patient’s diagnosis was studies have not addressed antibiotic abuse by pharmacists, made available to the pharmacists. who respond to the health-care needs of a substantial Implications of all the available evidence proportion of India’s population. Our findings suggest that non-adherence to regulatory standards Added value of this study is higher when the patient’s condition is unknown, and that We used standardised patients to quantify the extent of pharmacies prefer to treat in such cases rather than refer the antibiotic overuse in pharmacists for patients with tuberculosis. patient to appropriate care. These findings can inform We developed two standardised patient cases: first, a patient interventions to engage pharmacies in tuberculosis control and presenting with 2–3 weeks of pulmonary tuberculosis antimicrobial stewardship. of pharmacists to provide prescription-only drugs to levofloxacin, used in the treatment of tuberculosis) patients. Despite clear guidelines on the use of over-the- listed on a newly created Schedule H1. For H1 drugs, counter versus prescription-only drugs,10 enforcement is pharmacies require both a prescription from a qualified widely believed to be suboptimum.11,12 Pharmacies are medical practitioner and a separate register to record thought to be dispensing antibiotics and anti-tuberculosis the name and address of the prescriber, the patient, the drugs without prescriptions. Many tuberculosis patients names of the drugs and the quantity supplied.18 do seek medical advice and drugs from pharmacies,13 Schedule X, the most restrictive list, includes drugs driven by the ease of access and the possibility of avoiding such as narcotics, which require a prescription from a consultation charges by doctors.14 qualified provider to be retained by the retailer for Tuberculosis is a major problem in all three cities 2 years.19 studied (Delhi, Mumbai, and Patna), with notification We have previously assessed the quality of tuberculosis rates (officially reported) of 294, 210, and 77 per 100 000, care in India by health-care providers using standardised respectively.15 However, these rates are probably patients and use a similar method to study the practices underestimated because many cases treated in the of staff at pharmacies.20 Although standardised patients private sector are not notified.16 All three cities are are routinely used to assess pharmacy practices in low- experiencing rising rates of drug-resistant tuberculosis, income and high-income countries,21 to our knowledge, especially in the city of Mumbai,17 and it is widely believed no study has used standardised patients to assess that pharmacists are a key component of the dispensing pharmacy practices for tuberculosis in India. In our landscape and often a first contact for primary care. previous study, we validated the use of standardised Guidelines for pharmacies are specified under the patients for tuberculosis and showed the viability and Ministry of Health and Family Welfare’s Drugs and accuracy of this method for measuring quality of Cosmetics Rules Act, 1945.10 All antibiotics and steroids tuberculosis care along several dimensions, including are listed under two different schedules—Schedule H very low likelihood of detection, minimum to no study and Schedule H1. Schedule H drugs cannot be given to participation risk for either standardised patients or patients without a prescription from a qualified health-care providers, and high levels of accurate recall of medical practitioner. In 2013, regulations were further the clinical interaction among standardised patients. tightened, with anti-tuberculosis drugs (isoniazid, This study complements our previous validation study by rifampicin, ethambutol, and pyrazinamide) and extending the method to pharmacists. The method some fluoroquinolones (such as moxifloxacin and developed here addresses the dual objectives of, first, 1262 www.thelancet.com/infection Vol 16 November 2016 Articles assessment of pharmacists’ behaviour and drug use for a antibiotic use might be warranted for some of these patient with a complaint, but no prescription. Second, it conditions although not without a prescription from a allows us to assess how case management and drug use doctor. differs when the diagnosis is unknown versus confirmed. Standardised patients trained as Case 2 presented with 1 month of cough and fever and a tuberculosis-positive Methods laboratory report from a recent sputum smear test at a Study design and setting government dispensary. In this case, tuberculosis was This cross-sectional study was done in Delhi, Mumbai, confirmed, although the standardised patients, who and Patna. presented as uninformed patients, made it clear that they Through this multi-site study we aimed to assess the did not fully understand what the report said. In this medical advice and drug dispensing practices of situation, the pharmacist plausibly could know the correct pharmacies for standardised patients presenting with diagnosis and could recognise that short-term antibiotics either presumptive tuberculosis (Case 1) or micro- would not help, but also could realise that the patient biologically confirmed tuberculosis (Case 2). By assessing would still purchase antibiotics if offered because of their the difference in antibiotic use across the two cases for ignorance of the test results. Standardised patients did not the same pharmacists, we broke down the relative present with drug prescriptions; table 1 shows their importance of antibiotic misuse arising from the lack of opening statements and case scenarios. After each diagnosis (Case 1) versus antibiotic use despite a con- pharmacy visit, standardised patients were debriefed with firmed diagnosis for which antibiotics are contraindicated a structured questionnaire within 1 h of the visit. The (Case 2). To set the benchmark for what pharmacists accompanying appendix (pp 5, 6) provides more details on See Online for appendix should do when faced with such patients, we used the development of the cases and the recruitment and guidelines from the Government of India and the Indian characteristics of the standardised patients in the study. Pharmaceutical Association. These guidelines specify Cases are available from the authors by request. that pharmacies should counsel patients about tuber- culosis, identify and refer persons with tuberculosis Selection of pharmacies, standardised patient visits, symptoms to the nearest public health facilities for tuber- and study size culosis testing, and play a part in the provision of tuber- Standardised patients visited 54 pharmacies in Delhi culosis treatment.22 Therefore, pharmacists adhering to using a convenience sample from 28 low-income localities these guidelines should have referred the standardised in April, 2014. This phase of the study validated the patients to health-care providers without dispensing approach and provided key parameter estimates for power either antibiotics or steroids, both of which require a calculations employed in Mumbai and Patna. Based on prescription. these power calculations we sent standardised patients to 308 randomly sampled pharmacies in Mumbai and 260 in Standardised patients Patna between Nov 5, 2014, and Nov 29, 2015. 1200 (96%) The two cases of standardised patients used in our study of 1244 interactions were completed as planned, and we were adapted from our validation study in Delhi.15 completed both cases for a sampled pharmacy in 1156 Standardised patients trained as Case 1 presented with (93%) of 1244 scheduled interactions. The appendix 2–3 weeks of cough and fever and were directly seeking discusses the sample and sampling weights, case drugs from a pharmacy. Differential diagnosis for this development, standardised patient recruitment, sample case included upper respiratory tract infection, size calculations, drug identification, and deviations from pneumonia, asthma and acute or chronic bronchitis; the sampling scheme (pp 2–8). Case description Presentation of standardised patient Expected case management Case 1 Classic case of presumed Case 1 presents with the opening statement, “Sir, I have cough and fever that is not getting better. Please give Verbal or written referral to a DOTS tuberculosis with 2–3 weeks me some medicine.” At presentation, this case has had a 2–3 week cough, which occurred more during early centre or a health-care provider of cough and fever and morning and night, accompanied by a 2–3 week, on-and-off, low-grade fever. The patient was producing without dispensing any antibiotics directly seeking care from a sputum that did not contain any blood. The case would admit to a loss of appetite and to his or her clothes (including anti-tuberculosis drugs and chemist or pharmacist becoming a bit loose if prompted by the chemist. If the chemist asked about taking medicines for this illness, fluoroquinolones) or steroids the patient would say no Case 2 Chronic cough with a positive Case 2 presents with a positive sputum smear result visiting a chemist, presenting with the opening statement, Verbal or written referral to a DOTS sputum smear report for “Sir, I am having cough for nearly a month now and also have fever.” While showing a positive sputum report centre or a health-care provider tuberculosis from a to the chemist, the patient continues, “I went to the government dispensary and they asked me to get my without dispensing any antibiotics government dispensary and sputum tested. I have this report. Can you please give me some medicine?” At presentation, this case has had a (including anti-tuberculosis drugs and directly seeking care from a cough for 1 month and produces sputum without blood, accompanied by a 1 month, on-and-off, low-grade fluoroquinolones) or steroids chemist or pharmacist fever, which was more during evening times. Similar to Case 1, the case would admit to a loss of appetite and to his or her clothes becoming a bit loose if prompted by the chemist. If the chemist asked about taking medicines for this illness, the patient would say no DOTS=directly observed treatment, short-course. Table 1: Standardised patient case descriptions www.thelancet.com/infection Vol 16 November 2016 1263 Articles We obtained approvals from the ethics committees of Statistical analysis McGill University Health Centre in Montreal, Canada, and Our unit of analysis was a pharmacy-standardised patient the Institute of Socio-Economic Research on Development interaction irrespective of who (pharmacy owners, and Democracy (ISERDD) in New Delhi. Both ethics pharmacists, or pharmacy assistants) the standardised committees approved a waiver from obtaining informed patient interacted with. Whether the case was correctly consent from pharmacies in Mumbai and Patna. All managed was assessed from a tuberculosis perspective, individuals who participated as standardised patients were consistent with Standards for Tuberculosis Care in India hired as staff and trained to protect themselves from any and International Standards for Tuberculosis Care.7,8 We harmful medical interventions, such as avoiding injections, regarded ideal management for both cases as verbal or invasive tests, or consuming any drugs at the pharmacy. written referral to a health-care provider (public or private), without dispensing any antibiotics, including All cities (Delhi, Mumbai, and Patna) Patna and Mumbai only anti-tuberculosis drugs and fluoroquinolones, or steroids (table 1). Case 1 Case 2 Case 1 Case 2 We calculated the proportion and 95% CI for our Number of interactions 599 601 548 548 primary outcome, the proportion of interactions that Referral 96, 401, 75, 362, resulted in ideal management, as well as the proportion 0·16 (0·13–0·19) 0·67 (0·63–0·70) 0·14 (0·11–0·17) 0·66 (0·62–0·70) of interactions resulting in antibiotic, fluoroquinolone, Ideal case management 80, 372, 64, 335, and steroid use with appropriate sampling weights 0·13 (0·11–0·16) 0·62 (0·58–0·66) 0·12 (0·09–0·14) 0·61 (0·57–0·65) (appendix p 3). Drugs To assess the difference in case management and the Number of drugs 2·09 0·98 2·07 0·97 (1·99–2·20) (0·88–1·09) (1·97–2·18) (0·86–1·08) use of drugs across the two cases, we used a random Antibiotic 221, 98, 200, 88, intercept logit model with indicator variables for each 0·37 (0·33–0·41) 0·16 (0·13–0·19) 0·36 (0·32–0·41) 0·16 (0·13–0·19) city as additional controls. In view of the study design Steroid 45, 16, 37, 13, and since every sampled pharmacy was attempted by 0·08 (0·05–0·10) 0·03 (0·01–0·04) 0·07 (0·05–0·09) 0·02 (0·01–0·04) both cases, the choice of model (logit, logit with fixed Antibiotic or steroid 230, 104, 208, 94, effects, or logit with random intercepts) should have 0·38 (0·34–0·42) 0·17 (0·14–0·20) 0·38 (0·34–0·42) 0·17 (0·14–0·20) yielded similar unbiased estimates, with differences Fluoroquinolone 61, 23, 61, 23, arising only from the small portion of pharmacies that 0·10 (0·08–0·13) 0·04 (0·02–0·05) 0·11 (0·08–0·14) 0·04 (0·03–0·06) received one case but not the other. However, coefficients Schedule H 401, 188, 367, 172, 0·67 (0·63–0·71) 0·31 (0·28–0·35) 0·67 (0·63–0·71) 0·31 (0·27–0·35) from the random-intercepts model are more precisely Schedule H1 37, 19, 31, 16, estimated. The appendix (pp 10–13) provides a series of 0·06 (0·04–0·08) 0·03 (0·02–0·05) 0·06 (0·04–0·08) 0·03 (0·02–0·04) alternate estimates, with both marginal effects and odds Schedule X 0 0 0 0 ratios from different model specifications and confirm Anti-tuberculosis 0 0 0 0 that the results are very similar across specifications. All analyses were done using Stata (version 13). Data are n, proportion (95% CI) or mean (95% CI). Table 2: Management of Case 1 and Case 2 for all cities and for Patna and Mumbai only Role of the funding source The funders of the study had no role in study design, data collection, data analysis, data interpretation, or Odds ratio (95% CI) p value writing of the report. The corresponding author had full Ideal case management 21·0 (12·33–35·86) <0·0001 access to all the data in the study and had final Referral 16·40 (10·35–25·98) <0·0001 responsibility for the decision to submit for publication. Medication 0·05 (0·03–0·09) <0·0001 Antibiotic 0·21 (0·15–0·31) <0·0001 Results Fluoroquinolone 0·31 (0·18–0·53) <0·0001 96 (16%) of 599 pharmacies (95% CI 13–19) referred Schedule H 0·15 (0·11–0·21) <0·0001 Case 1 interactions to health-care providers, but because Schedule H1 0·44 (0·23–0·82) 0·0099 in 16 (17%) of these 96 cases the standardised patient was Steroid 0·27 (0·14–0·53) 0·0001 also given an antibiotic or steroid (95% CI 11–25), ideal 0·01 0·1 1·0 10·0 100·0 case management (referral to a health-care provider without any antibiotics and steroids) occurred in 80 (13%) Favours Case 1 Favours Case 2 of 599 Case 1 interactions (95% CI 11–16). Overall, Figure 1: Odds ratios for case management outcomes for Case 1 versus Case 2 antibiotics were used in 221 (37%; 95% CI 33–41) of Reported odds ratios are from a random-intercepts model using each pharmacy as its own control, with city fixed 599 interactions, steroids in 45 (8%; 95% CI 6–10), and effects. Odds ratios greater than 1 favour Case 2. Referral is any instance in which the pharmacy staff recommended fluoroquinolones in 61 (10%; 95% CI 8–13). Because that the standardised patient seeks further care from a health-care provider. Ideal case management for both cases is defined as a referral without the dispensing of antibiotics or steroids. Schedule H, H1, and X drugs are defined as Schedule H drugs also include prescription-only drugs per the Drugs and Cosmetics Act, 1945, of the Ministry of Health and Family Welfare, Government of India and its that are not antibiotics or steroids (eg, ibuprofen or amendments. cetirizine), the use of these drugs was higher (401 [67%] of 1264 www.thelancet.com/infection Vol 16 November 2016 Articles A No referral B Referral Delhi 100 Referral Mumbai 5% Patna 12% 7% 5% Ideal case management Percentage of interactions (%) 75 34% 32% Antibiotic 71% 80% 2% Steroid 50 3% Antibiotic or steroid 53% 49% 3% 25 14% Fluoroquinolone 6% 13% 14% Schedule H 0 Case 1 Case 2 Case 1 Case 2 (503/599) (200/601) (96/599) (401/601) Schedule H1 No medication Steroid Antibiotic and steroid No antibiotic or steroid Schedule X Antibiotic Figure 2: Drug use by referral decisions for two standardised patient cases Anti-tuberculosis medicine Each panel describes the use of drugs in each case; the first shows pharmacies that did not refer the standardised patient to another health-care provider (left 0 25 50 75 100 panel) and the second shows those who did (right panel). Both cases are Percentage of interactions (%) presented in percentages; the percentages making referral decisions are shown below the case labels in each panel. Percentages indicate the number of Figure 3: Management of both Case 1 and Case 2 combined by city interactions within each case-referral category dispensing the indicated types of Referral is any instance in which the pharmacy staff recommended that the drugs; percentages may add to more than 100% due to rounding. standardised patient seek further care from a health-care provider. Ideal case management for both cases is defined as a referral without the dispensing of antibiotics or steroids. Schedule H, H1, and X drugs are defined as per the Drugs and Cosmetics Act, 1945, of the Ministry of Health and Family Welfare, 599 interactions, 95% CI 63–71). The use of Schedule H1 Government of India and its amendments. drugs was notably lower (37 [6%] of 599, 95% CI 4–8) and Schedule X drugs and anti-tuberculosis drugs were never given. Table 2 provides the mean proportions of the key p<0·0001) and for fluoroquinolones 0·31 (0·18–0·53; outcome variables in all cities combined for Case 1 and p<0·0001). We also note that of the 497 referrals across Case 2. Since the sampling scheme was different for the two cases, 301 (60%) were to doctors in the private Delhi compared with Mumbai and Patna, we also provide sector and the remaining 40% were to the public sector results excluding Delhi (table 2), and for each city by case (data not shown). In only three instances was the (appendix pp 8,9). standardised patient referred specifically to a directly By contrast with Case 1, 401 (67%) of 601 pharmacies observed treatment, short-course (DOTS) centre. (95% CI 63–70) referred Case 2 to a health-care provider In terms of behaviour conditional on referral, the (table 2). As before, some patients received antibiotics or differences between Case 1 and Case 2 reflect, to a steroids even with a referral, so ideal case management substantial degree, the large increase in referrals for was recorded in 372 (62%) of 601 interactions (95% CI Case 2. Figure 2 shows the proportion of interactions that 58–66). Antibiotics, steroids, and fluoroquinolones were received antibiotics or steroids, or both, or no drug all used much less frequently, although Schedule H separated by case and referral decision. Both for Case 1 drugs were still given in 188 (31%) of 601 interactions and Case 2, the use of antibiotics or steroids and the total (95% CI 28–35). As before, Schedule X and anti- number of drugs fell when the pharmacist referred the tuberculosis drugs were never used. patient (0·75 for Case 1, 95% CI 0·48–1·02 vs 0·38 for Figure 1 uses the random-intercept model together Case 2, 0·29–0·46; data not shown). However, with indicator variables for each city to estimate the conditioning on the decision to refer, the difference in difference in pharmacy behaviour for the main outcome pharmacist behaviour across the two cases was much variables as odds ratios. All these differences were smaller. significant and precisely estimated. For instance, the The practice of pharmacies varied across cities, adjusted odds of pharmacies referring a standardised although caution is warranted in interpreting these patient with a sputum smear-positive tuberculosis report results in view of the different sampling methods used to a health-care provider without dispensing antibiotics (appendix p 3). We noted similar patterns across the and steroids (ideal case management) was 21·03 (95% CI three cities of high use of Schedule H drugs, referrals, 12·33–35·86; p<0·0001) for Case 2 relative to Case 1; the and ideal case management (figure 3). Two differences odds ratio for antibiotic use was 0·21 (0·15–0·31; worth highlighting are that compared with Mumbai, the www.thelancet.com/infection Vol 16 November 2016 1265 Articles dispensed fluoroquinolones (eg, ciprofloxacin, levo- A Case 1 (n=599) classic case of B Case 2 (n=601) tuberculosis case presumed tuberculosis with positive sputum report floxacin, ofloxacin), whereas 45 (8%) of 599 gave steroids Ayurvedic (33) Ayurvedic (16) such as betamethasone and prednisolone (95% CI 6–10). Herbal (10) Amoxicillin (50) For Case 2, pharmacies dispensed 0·98 drugs on average Amoxicillin (100) Azithromycin (16) (95% CI 0·88–1·09). The classes of drugs dispensed for Ofloxacin (25) Cefixime (11) Case 2 were similar to Case 1, although the overall Ciprofloxacin (24) Levofloxacin (7) frequencies were much lower. This finding is again Azithromycin (23) Ofloxacin (7) consistent with the result that the difference in behaviour Cefixime (19) Ciprofloxacin (5) between the two cases was driven, to a large extent, by the Levofloxacin (14) Paracetamol (135) sharp increase in referrals for Case 2. Ampicillin (8) Nimesulide (54) Roxithromycin (8) Diclofenac (20) Discussion Cloxacillin (6) Chlorzoxazone (9) To our knowledge, this is the first study that used Erythromycin (6) Ranitidine (6) Betamethasone (8) standardised patients to examine how pharmacies in Paracetamol (319) Nimesulide (123) Prednisolone (5) India treat patients with tuberculosis symptoms and Diclofenac (64) Chlorpheniramine (43) diagnosed tuberculosis, complementing our recent study Aceclofenac (10) Guaiphenesin (39) that assessed tuberculosis management by health-care Chlorzoxazone (10) Phenylephrine (32) providers.20 Because the standardised patient method Ranitidine (14) Cetirizine (28) standardises the presentation of the underlying condition Omeprazole (6) Phenylpropanolamine (7) across different providers,23 the results are reliable, valid, Betamethasone (25) Levocetirizine (6) and comparable across pharmacies. The similar patterns Prednisolone (11) Vitamins (11) we recorded across the three cities suggest that the Dexamethasone (9) Ammonium chloride (53) results might be generalisable to other urban areas in Chlorpheniramine (98) Bromhexine (50) India. Guaiphenesin (84) Dextromethorphan (47) A key finding is that none of the pharmacies in our Phenylephrine (57) Diphenhydramine (39) study dispensed first-line anti-tuberculosis drugs. Cetirizine (42) Sodium citrate (39) Levocetirizine (11) Ambroxol (7) Concerns regarding the use of anti-tuberculosis drugs by Phenylpropanolamine (8) Terbutaline (10) pharmacies seem to be unfounded, at least in major Vitamins (17) cities, and pharmacies are unlikely sources of irrational 0 20 40 60 Bromhexine (91) drug use that contributes to multidrug-resistant tuber- Percentage of interactions (%) Dextromethorphan (89) culosis. Why pharmacists do not dispense tuberculosis Sodium citrate (76) drugs requires further research, but the fact that Diphenhydramine (74) tuberculosis drugs (unlike antibiotics such as amoxicillin) Ambroxol (19) Ayurvedic Anti-allergy are considered toxic and that tuberculosis requires long- Terbutaline (29) Antibiotics Other term treatment might play a part. Proactiveness of the Analgesics Cough syrups 0 20 40 60 Antiulcer Bronchodilators Indian National Tuberculosis Control Program in Percentage of interactions (%) Steroids including tuberculosis drugs under Schedule H1 and the requirement to document tuberculosis drug prescriptions Figure 4: Active ingredients in drugs given for each case might also have reduced abuse. The frequency with which each listed active ingredient was contained in drugs given to standardised patients for each case. The number in brackets is the number of interactions in which that active ingredient was recorded. However, our findings showed that 38% of the pharmacies dispensed antibiotics or steroids to people use of antibiotics, steroids, fluoroquinolones, and with tuberculosis symptoms but no test results. The use Schedule H1 drugs were all much higher in Patna; and of fluoroquinolones in 7% and steroids in 5% of that there was no fluoroquinolone use in Delhi and little interactions is especially worrying because these drugs use in Mumbai compared with Patna. These differences delay tuberculosis diagnosis.5,24 Additionally, fluoro- are robust to adjustment for differences in the quinolones are also an essential part of multidrug- standardised patients used across different cities, an resistant tuberculosis treatment regimens and emerging analysis that we did by comparing outcomes only among regimens, so quinolone abuse is a concern.5 the (smaller) group of standardised patients who were The widespread use of antibiotics and steroids for common to two or more cities (appendix pp 13,14). respiratory symptoms also has implications for com- In terms of type of drugs dispensed, for Case 1, munity-acquired infections more generally. Unnecessary pharmacies dispensed 2·09 drugs on average (95% CI use of fluoroquinolones is a major risk factor for creating 1·99–2·20; figure 4). The most common classes of drugs highly resistant Gram-negative enteric bacteria (eg, dispensed were analgesics such as paracetamol and extended spectrum beta-lactamase resistance) that might nimesulide, antibiotics, cough syrups, and anti-allergy cause diarrhoeal illness, bacteraemia, and other drugs. Among antibiotics, amoxicillin was the most infections, especially in India.25 The common use of common, and 61 (10%) of 599 (95% CI 8–13) pharmacies aminopenicillins (eg, amoxicillin) and macrolides (eg, 1266 www.thelancet.com/infection Vol 16 November 2016 Articles azithromycin) for respiratory symptoms identified in our appendix p 9). Fourth, differences between Case 1 and study might contribute to resistant strains of common Case 2 could reflect variation in the standardised patient respiratory pathogens such as Streptococcus pneumoniae profile. Because different standardised patients were and Haemophilus influenzae.26 In addition to potentially assigned to the two cases with no crossover, we cannot delaying tuberculosis diagnosis, unnecessary use of assess this possibility. Generally, the inclusion of steroids is associated with an increased risk of developing standardised patient characteristics has little effect on lower respiratory tract infection, cellulitis, herpes zoster, estimated coefficients in previous standardised patient and candidiasis.27 studies and our coefficients remain stable when we Our results also clearly show that a first-order problem account for standardised patient sex, height, and weight both in the management of tuberculosis and anti- (appendix pp 13, 14). microbial resistance is the information that patients To conclude, our study adds to the growing evidence in present to the pharmacist. Confirmed diagnoses India on antibiotic abuse, but also underscores that the discipline what pharmacists do, with sharp increases in use of antibiotics is mediated by drug category and the ideal management and large decreases in antibiotic use. information that patients present. Although antibiotic This dramatic difference suggests that the main use is high and such use can delay diagnosis, none of the challenge faced by pharmacists is confusion about the pharmacies dispensed anti-tuberculosis drugs and the likely diagnosis, in which case better training regarding use of stronger fluoroquinolone antibiotics and heavily tuberculosis symptoms and encouraging early referrals restricted drug classes was low. Furthermore, the use of for patient with tuberculosis symptoms might help. all antibiotics decreased sharply when the patient’s Lastly, our study shows the value of the standardised diagnosis was revealed to the pharmacists. These patient method in tracking inappropriate antibiotic use.28 findings can inform interventions to engage pharmacies Although prescription audits can be used, prescriptions in tuberculosis control and antimicrobial stewardship. do not capture the off-prescription use of drugs and often Contributors do not include diagnoses. JD and MP obtained funding and designed the study. JD, AK, SS, VD, Although the behaviour change in Case 2 suggests and MP developed the standardised patient cases and scripts. AK and RKD collected data and supervised data collection. BD, SS, and RS that pharmacists substantially decrease the use of coded the data. VD, MP, and AK trained the standardised patients. unnecessary drugs when the diagnosis is known, it is SS, JD, AK and BD analysed the data. SS, JD, BD, AK, RS, AM, and unknown why some pharmacists give antibiotics and MP interpreted the data. The report was written by SS, JD, BD, MP, others do not; neither can we uncover the reasons why AK, and SB, and all authors provided critical review and comments to the revision of the report. pharmacists are unwilling to follow regulations regarding drug use in these three cities. It is unclear Declaration of interests MP serves as a consultant for the Bill & Melinda Gates Foundation. whether the variation in our data is explained by the He has no financial conflicts to disclose. All other authors declare no competence and qualification of the person providing competing interests. advice in pharmacies, which we did not track in the Acknowledgments study. Qualitative evidence suggests that a combination This study is funded by Grand Challenges Canada (grant ID: S5 0373-01), of other factors might also be at play, including pharma- Bill & Melinda Gates Foundation (grant number: OPP1091843), ceutical industry marketing techniques, business Knowledge for Change Program, World Bank Development Research Group. SS is supported by a fellowship from the Canadian Thoracic models followed by local providers, and active demand Society and is also a senior operations research fellow at Center for from patients for medicines.11,29 Pharmacists in Delhi Operational Research, The Union (Paris, France). RS is supported by a have described overstock, near-expiry, and undersupply Fogarty Global Health Equity Scholars Fellowship (NIAID R25 TW009338). JD and BD received funds from the Knowledge for Change as further factors precipitating misuse of antibiotics Program (The World Bank). MP is a recipient of Tier 1 Canada Research and restricted drugs.11 Chair from Canadian Institutes of Health Research. We thank Second, we noted significantly higher use of antibiotics Puneet Dewan, Sarang Deo, Nim Pathy, and Vaibhav Saria for useful and quinolones in Patna than in Mumbai pointing to input on analysis and interpretation; Rajan Singh, Purshottam, Chinar Singh, Geeta, Devender, Varun Kumar, Anand Kumar, Babloo, some differences across cities. We are able to rule out and Charu Nanda who supervised and implemented the ISERDD that these differences reflect the composition of fieldwork; and all the standardised patients for their dedication and hard standardised patients deployed across cities (appendix work. The findings, interpretations, and conclusions expressed here are p 13), but with an effective sample size of only three those of the authors and do not necessarily represent the views of the World Bank, its Executive Directors, or the governments they represent. cities, we cannot explain this variation. Also, our study does not provide evidence on how pharmacists in rural References 1 Shallcross LJ, Davies SC. The World Health Assembly resolution on areas manage patients with tuberculosis or tuberculosis antimicrobial resistance. J Antimicrob Chemother 2014; 69: 2883–85. symptoms. Third, our study reflects what happens when 2 Van Boeckel TP, Gandra S, Ashok A, et al. Global antibiotic pharmacists receive a completely unknown patient as consumption 2000 to 2010: an analysis of national pharmaceutical sales data. Lancet Infect Dis 2014; 14: 742–50. opposed to a known, regular client, or a client who 3 Levy SB, Marshall B. Antibacterial resistance worldwide: causes, returns to the pharmacist after one round of ineffective challenges and responses. Nat Med 2004; 10: S122–S9. treatment. We note though that only 5–6% of pharmacists 4 Laxminarayan R, Chaudhury RR. Antibiotic resistance in India: asked the patient to return (if they did not feel better; drivers and opportunities for action. PLoS Med 2016; 13: e1001974. www.thelancet.com/infection Vol 16 November 2016 1267 Articles 5 Chen T-C, Lu P-L, Lin C-Y, Lin W-R, Chen Y-H. 18 Government of India Ministry of Health and Family Welfare. Fluoroquinolones are associated with delayed treatment and Notification G.S.R. 588(E). In: Health Do, ed. New Delhi: resistance in tuberculosis: a systematic review and meta-analysis. The Gazette of India Extraordinary, 2013. Int J Infect Dis 2011; 15: e211-e6. 19 Government of India Ministry of Health and Family Welfare. 6 Zhang Y, Yew W. Mechanisms of drug resistance in Mycobacterium National Formulary of India 4th Edition. New Delhi: Indian tuberculosis [State of the art series. Drug-resistant tuberculosis. Pharmacopoeia Commission, 2011. Edited by CY. Chiang. Number 1 in the series]. Int J Tuberc Lung Dis 20 Das J, Kwan A, Daniels B, et al. Use of standardised patients to 2009; 13: 1320–30. assess quality of tuberculosis care: a pilot, cross-sectional study. 7 TB Care I. International standards for tuberculosis care, edn 3. Lancet Infect Dis 2015; 15: 1305–13. The Hague: TB CARE I , 2014. 21 Xu T, de Almeida Neto AC, Moles RJ. A systematic review of 8 WHO Country Office for India. Standards for tuberculosis Care in simulated-patient methods used in community pharmacy to assess India. Geneva: World Health Organization, 2014. the provision of non-prescription medicines. Int J Pharm Pract 9 All India Organisation of Chemists and Druggists. All India 2012; 20: 307–19. Organisation of Chemists & Druggists history. 2015. http://www. 22 Central TB Division, Ministry of Health and Family Welfare, aiocd.net/ (accessed Sept 24, 2015). Government of India and Indian Pharmaceutical Association. 10 Ministry of Health and Family Welfare, Government of India. Revised National Tuberculosis Control Programme training module The Drugs and Cosmetics Rules, 1945 [As amended up to 15th for community pharmacists. 2013. http://www.pci.nic.in/Circulars/ August, 2013]. http://www.mohfw.nic.in/WriteReadData/ Training%20Module.pdf (accessed Aug 15, 2016). l892s/43503435431421382269.pdf (accessed Aug 15, 2016). 23 Das J, Holla A, Das V, Mohanan M, Tabak D, Chan B. In urban and 11 Kotwani A, Wattal C, Joshi P, Holloway K. Irrational use of rural India, a standardized patient study showed low levels of antibiotics and role of the pharmacist: an insight from a qualitative provider training and huge quality gaps. Health Aff 2012; study in New Delhi, India. J Clin Pharm Ther 2012; 37: 308–12. 31: 2774–84. 12 Salunkhe S, Pandit V, Dawane J, Sarda K, More C. Study of over the 24 Cisneros JR, Murray KM. Corticosteroids in tuberculosis. counter sale of antimicrobials in pharmacy outlets in Pune, India: Ann Pharmacother 1996; 30: 1298–303. a cross sectional study. Int J Pharm Bio Sci 2013; 4: 616–22. 25 Kantele A, Laaveri T, Mero S, et al. Antimicrobials increase 13 Kapoor SK, Raman AV, Sachdeva KS, Satyanarayana S. How did the travelers’ risk of colonization by extended-spectrum tuberculosis patients reach DOTS services in Delhi? A study of betalactamase-producing Enterobacteriaceae. Clin Infect Dis 2015; patient treatment seeking behavior. PLoS One 2012; 7: e42458. 60: 837–46. 14 Goel P, Ross-Degnan D, Berman P, Soumerai S. Retail pharmacies 26 Torumkuney D, Chaiwarith R, Reechaipichitkul W, et al. in developing countries: a behavior and intervention framework. Results from the survey of antibiotic resistance (SOAR) 2012–14 in Soc Sci Med 1996; 42: 1155–61. Thailand, India, South Korea and Singapore. 15 Central Tuberculosis Division. Tuberculosis India 2014: annual J Antimicrob Chemother 2016; 71 (suppl 1): i3–i19. status report of the Revised National Tuberculosis Control 27 Fardet L, Petersen I, Nazareth I. Common infections in patients Programme. Minstry of Health and Family Welfare, Government prescribed systemic glucocorticoids in primary care: of India, 2014. a population-based cohort study. PLoS Med 2016; 13: e1002024. 16 Pai M, Dewan P. Testing and treating the missing millions with 28 Ghafur A, Mathai D, Muruganathan A, et al. The Chennai tuberculosis. PLoS Med 2015; 12: e1001805. Declaration: a roadmap to tackle the challenge of antimicrobial 17 Udwadia ZF, Amale RA, Ajbani KK, Rodrigues C. resistance. Indian J Cancer 2013; 50: 71–73. Totally drug-resistant tuberculosis in India. Clin Infect Dis 2012; 29 Kamat VR, Nichter M. Pharmacies, self-medication and 54: 579–81. pharmaceutical marketing in Bombay, India. Soc Sci Med 1998; 47: 779–94. 1268 www.thelancet.com/infection Vol 16 November 2016