Proxy indicators to estimate appropriateness of antibiotic prescriptions by general practitioners: a proof-of-concept cross-sectional study based on reimbursement data, north-eastern France 2017

Background In most countries, including France, data on clinical indications for outpatient antibiotic prescriptions are not available, making it impossible to assess appropriateness of antibiotic use at prescription level. Aim Our objectives were to: (i) propose proxy indicators (PIs) to estimate appropriateness of antibiotic use at general practitioner (GP) level based on routine reimbursement data; and (ii) assess PIs’ performance scores and their clinimetric properties using a large regional reimbursement database. Methods A recent systematic literature review on quality indicators was the starting point for defining a set of PIs, taking French national guidelines into account. We performed a cross-sectional study analysing National Health Insurance data (available at prescriber and patient levels) on antibiotics prescribed by GPs in 2017 for individuals living in north-eastern France. We measured performance scores of the PIs and their case-mix stability, and tested their measurability, applicability, and room for improvement (clinimetric properties). Results The 3,087 GPs included in this study prescribed a total of 2,077,249 antibiotic treatments. We defined 10 PIs with specific numerators, denominators and targets. Performance was low for almost all indicators ranging from 9% to 75%, with values < 30% for eight of 10 indicators. For all PIs, we found large variation between GPs and patient populations (case-mix stability). Regarding clinimetric properties, all PIs were measurable, applicable, and showed high improvement potential. Conclusions The set of 10 PIs showed satisfactory clinimetric properties and might be used to estimate appropriateness of antibiotic prescribing in primary care, in an automated way within antibiotic stewardship programmes.


Proxy indicator (PI)
Based on the following quality indicator (QI), identified in the literature review + consensus procedure a French national guidelines/recommendations used to adapt the definition of the QI to the French context French or international guidelines/recommendations used to set the target (optimal/acceptable) PI 1: Antibiotic prescriptions against UTI in men (ratio) OQI-4 Some antibiotics should be rarely prescribed -National guidelines on Urinary Tract Infections (UTI) 1,2 -Nitrofurantoin, fosfomycin-trometamol, 1st-generation quinolones (J01MB), norfloxacin, enoxacin and lomefloxacin are not recommended in male UTIs -The optimal target is 0, but we also set an acceptable target at <0.5, based on expert opinion (since guidelines are not applicable to all patients and some of these antibiotics might be used as lastresort treatments, e.g. in complex patients with relapsing UTIs or UTIs due to multi-drug resistant bacteria) PI 2: Antibiotic prescriptions against UTI in women (ratio) OQI-3 Outpatients should receive antibiotic therapy compliant with guidelines -National guidelines on Urinary Tract Infections (UTIs) 1,2 -Cystitis are much more frequent in primary care as compared to pyelonephritis. Regarding cystitis in women, fluoroquinolones are never first-line treatments (whereas this might be the case for empirical treatment for pyelonephritis). In female cystitis, nitrofurantoin, pivmecillinam and fosfomycin-trometamol are the usual 1 st -line treatments (depending on the clinical situation). Moreover, nitrofurantoin, pivmecillinam and fosfomycin-trometamol are exclusively recommended for UTIs. This is not the case for fluoroquinolones, which can be indicated in very selected indications (e.g. levofloxacin is recommended as second-line treatment in some respiratory infections), but the level of use of fluoroquinolones should be very limited if guidelines are complied with. 3 -For all these reasons, we selected a ratio >1, based on expert opinion, even though the optimal ratio might be even higher PI 3: Repeated prescription of quinolones (%) OQI-3 Outpatients should receive antibiotic therapy compliant with guidelines -National guidelines on Urinary Tract Infections (UTI) and lower respiratory tract infections 1,2 -National recommendations on fluoroquinolone use 3 -Based on these recommendations, a quinolone should not be used (whenever possible) among patients having been prescribed a quinolone in the preceding 6 months -The optimal target is 0, but we also set an acceptable target at <10%, based on expert opinion (since guidelines are not applicable in all cases) A target 5% is used in Scotland as a national indicator. 8 We also set an acceptable target at <10%, based on expert opinion and the range of existing European data. 7

PI 6:
Amoxicillin / second-line antibiotics prescriptions (ratio) OQI-3 Outpatients should receive antibiotic therapy compliant with guidelines -National guidelines on the most common infections encountered in primary care 1,2 -In France, 70% of all antibiotics prescribed in primary care concern upper and lower respiratory tract infections. 9 -Amoxicillin is the 1 st -line recommended treatment for almost all bacterial respiratory tract infections, whereas co-amoxiclav, cephalosporins and macrolides are usually 2 nd -line treatments -For all these reasons, we selected a ratio >1, based on expert opinion, even though the optimal ratio might be even higher PI 7: Prescription of not indicated antibiotics (%) OQI-4 Some antibiotics should be rarely prescribed -National guidelines on the most common infections encountered in primary care 1,2 -Lomefloxacin, moxifloxacin, 1st-generation quinolones (J01MB), norfloxacin, enoxacin and lomefloxacin, telithromycin, spiramycin-metronidazole and cefaclor are not indicated according to these national guidelines -The optimal target is 0, but we also set an acceptable target at <0.5, based on expert opinion (since guidelines are not applicable to all patients and some of these antibiotics might be used as lastresort treatments) OQI-3 Outpatients should receive antibiotic therapy compliant with guidelines -National guidelines on the most common infections encountered in primary care 1,2 -The maximum recommended duration of antibiotic treatment for almost all bacterial infections encountered in GP practice is one week, and durations <1 week are common -We selected here antibiotics that are almost exclusively recommended in respiratory tract and skin infections, or cystitis, where recommended durations <8 days are the rule -We set an optimal target at <5% for durations of 9 days or more, and an acceptable target at <10%, based on expert opinion -In France, unit dispensing is not in place, so antibiotics are delivered using packages. The packages' size is however adapted to the most frequent durations advised in national recommendations -See supplementary Table 2 for more details regarding the calculation of this indicator PI 9: Co-prescription of antibiotic and non-steroidal antiinflammatory drugs (NSAIDs) (%) OQI-3 Outpatients should receive antibiotic therapy compliant with guidelines -NSAIDs are never indicated and should be avoided in bacterial infections encountered in primary care, except for fever in children not controlled by paracetamol. 10 -National guidelines on the most common infections encountered in primary care 1,2 -This indicator does not take into account self-medication with NSAIDs (some NSAIDs can be bought at the patient's cost in community pharmacies without a prescription, and are therefore not recorded in the NHI database) -The optimal target is 0, but we also set an acceptable target at <5%, based on expert opinion (since guidelines are not applicable to all patients, and some patients might be on NSAIDs for another reason) PI 10: Co-prescription of antibiotic and corticosteroids (%) OQI-3 Outpatients should receive antibiotic therapy compliant with guidelines -Corticosteroids are never indicated and should be avoided in bacterial infections encountered in primary care, except for blocked acute bacterial sinusitis (which is very rare and usually managed by ENT specialists) -National guidelines on the most common infections encountered in primary care 1,2 -The optimal target is 0, but we also set an acceptable target at <5%, based on expert opinion (since guidelines are not applicable to all patients, and some patients might be on corticosteroids for another reason) a J Antimicrob Chemother. 2018;73(suppl_6):vi40-vi49

Supplementary Table S2. Detailed procedure for the calculation of Proxy Indicator n°8: Estimated duration of antibiotic prescriptions > 8 days
In France, unit dispensing is not in place, so antibiotics are delivered using packages. The packages' sizes available on the French market should however be adapted by pharmaceutical companies to the most frequent durations advised in national recommendations; an experiment conducted in 100 community pharmacies in France showed that per-unit dispensing reduced by 10% the number of pills supplied. 11 To account for this excess of dispensed pills as compared to the exact duration prescribed by the GP, we set a target at 9 days or more of treatment, instead of 8 days or more (see Supplementary Table S1 for the rationale for 8 days).
The French National Health Insurance database does not contain information on the precise prescribed daily dose or duration, but only on the dispensed packages. We here calculated the quantity of dispensed antibiotics for each prescription (number of packages multiplied by the quantity of antibiotic [in grams] per package).
We then determined a 'usually recommended total daily dose' for the most frequently prescribed antibiotics (see details in the Table below). This usual total daily dose is the most commonly prescribed daily dose in general practice in adults, according to national guidelines. 1,2 We did not plan initially to restrict this metric to adults > 16 years old, since: (i) GPs usually take care of few children; (ii) the total daily dose recommended in children > 40-50 kg is usually the same as for adults; and (iii) we planned to explore differences according to age classes in our case-mix analyses.
We excluded the following antibiotics: (i) azithromycin and fosfomycin-trometamol, since both are recommended for short durations given their long half-life; and (ii) doxycycline and fluoroquinolones since these antibiotics might be quite frequently prescribed for durations of more than one week according to national recommendations.
We acknowledge that this indicator only approximates the real duration prescribed by the general practitioner (GP), but data on days-of-therapy are not available in routine in existing French databases. We might therefore: overestimate the real duration prescribed by the GP if the quantity dispensed exceeds the quantity prescribed by the GP (in case the packages do not match the exact quantity prescribed) or if the total daily dose prescribed by the GP exceeds the usual prescribed total daily dose considered here (see Table below) -underestimate the real duration if the total daily dose prescribed by the GP is lower than the usual prescribed total daily dose considered here (see Table below