Diagnosis-linked antibiotic prescribing in Swedish primary care - a comparison between in-hours and out-of-hours

The rise of antibiotic resistance is a global public health threat according to the World Health Organization [1], and antibiotic overuse is common and results in medicalization, unnecessary costs, and increased antibiotic resistance [2]. However, studies on antibiotic prescribing in primary care regardless of indication show a high level of variability between physicians in different countries [3‐5].

In primary care in-hours (IH) are usually office hours (in Sweden 08:00 to 17:00) during business days, and out-of-hours (OOH) are the remaining hours. Earlier studies of OOH care have suggested that compared to IH care there are lower adherence to antibiotic guidelines [6, 7], a higher antibiotic prescribing rate [8, 9], a higher rate of prescriptions for broad-spectrum antibiotics [8], and more antibiotic prescriptions during weekends than weekday evenings [10]. In a qualitative study from Belgium, the physicians reported that the threshold for prescribing antibiotics was lower during OOH, but the choice of antibiotics was the same [11]. A more recent Belgian OOH study showed a high antibiotic prescribing rate for all indications, a high rate of not using recommended antibiotics, and an overuse of quinolones [12]. However, a Dutch study found the prescribing quality to be appropriate, and the higher rates of prescribing in OOH were explained by a different population of presenting patients [13]. No previous study has compared the OOH treatment of infections to IH within the same population.

Although Sweden belongs to the European countries with low levels of antibiotic prescriptions, there is still room for improvement [14]. Previous registry-based studies in Sweden have shown a significant reduction in antibiotic prescriptions over the last decade, but these studies have not included OOH [15‐17]. Several Swedish national guidelines concerning the evaluation and treatment of infectious diseases have been published [18‐22], and generally these guidelines aim at better diagnostics, fewer antibiotics, and more targeted treatments.

Because visits for infectious diseases are common at OOH centres, it is important to evaluate whether OOH visits are associated with increased antibiotic prescribing rates because this would warrant interventions in OOH settings.

The purpose of the study was to describe the trends in antibiotic prescribing over time and to compare diagnosis-linked prescribing in general and in detail between IH and OOH in the same population.

The physician visit rate for infections varied during the study and reached a maximum of 469 visits per 1000 inhabitants per year in 2011 and a minimum of 398 visits in 2014. Female patients have more infection visits than male patients, 502 and 366 visits per 1000 inhabitants per year respectively. Children 0–4 years and adults over 80 years had the highest visit rates, 995 and 576 visits per 1000 inhabitants per year respectively. No significant trends were observed in total visit rate nor in visit rate by sex, but the mean annual change in visit rate per 1000 inhabitants per year decreased in children 0–4 years (− 33.7 (95% CI − 56.0 to − 11.5)), increased in adults 65–79 years (7.7 (95% CI 1.1 to 14.3) and in adults over 80 years (13.9 (95% CI 7.6 to 20.2)) (Tables 1 and 2).

The antibiotic prescriptions per 1000 inhabitants per year decreased significantly from 266 prescriptions in 2006 to 194 prescriptions in 2014 (mean annual change − 8.5 (95% CI − 11.9 to − 5.2)). There was no sex difference, but the decrease in antibiotic prescriptions was more pronounced in children 0–4 years (mean annual change − 35.2 (95% CI − 46.9 to − 23.5)) and in children 5–19 years (mean annual change − 11.7 (95% CI − 17.0 to − 6.5). The antibiotic prescribing frequency decreased mainly for RTIs (mean annual change − 6.5 (95% CI − 9.0 to − 3.9)), explaining 76% of the total reduction. Antibiotic prescriptions without an infection diagnosis and prescriptions for UTIs also decreased, explaining a further 11 and 8% of the total reduction, respectively (Tables 3 and 4).

Of all antibiotic prescriptions, 75% were linked to an infection visit on the same day, another 3% were linked to an infection visit within a week before the prescription day, and finally 22% were not possible to link to an infection visit. These proportions were stable during the study period. Of all antibiotics prescribed at visits, 66% were antibiotics commonly used for RTIs, 12% were commonly used for SSIs, 16% were commonly used for UTIs, and 6% were other antibiotics. Of the antibiotics prescribed without an infection diagnosis, 38% were antibiotics commonly used for RTIs, 25% were commonly used for SSIs, 29% were commonly used for UTIs, and 8% were other antibiotics. Of the UTI antibiotics, 36% were prescribed without an infection diagnosis.

During the study period, the OOH infection visits decreased from 65 visits per 1000 inhabitants in 2006 to 43 visits in 2014 (mean annual change − 3.0 visits (95% CI − 4.2 to − 1.7)). Also, the antibiotic prescribing decreased from 43 prescriptions per 1000 inhabitants in 2006 to 26 prescriptions in 2014 (mean annual change − 2.2 prescriptions (95% CI − 3.3 to − 1.2)).

The diagnoses and antibiotic prescription rates between IH and OOH are shown in Table 5. During IH, there were 382 infection visits per 1000 inhabitants per year compared to 51.4 during OOH. Thus 12% of all visits were during OOH. RTIs were the most common diagnoses during both IH and OOH. However, acute otitis media, pharyngotonsillitis, and lower UTIs were more common during OOH. A total of 15% of all antibiotics were prescribed during OOH. The likelihood of receiving an antibiotic prescription was 55% during OOH visits compared to 41% during IH visits. The unadjusted relative risk of antibiotic prescribing in OOH was 1.37 (95% CI 1.36 to 1.38) compared to IH. The difference remained unchanged when only adjusted for age and sex 1.37 (95% CI 1.37 to 1.38) and 1.37 (95% CI 1.37 to 1.38), respectively. However, when adjusted for age, sex, and diagnosis the relative risk of antibiotic prescribing during OOH was 1.09 (95% CI 1.08 to 1.10) compared to IH. No difference was found between the two OOHCs. Age and sex adjusted relative risks of antibiotic prescribing during OOH per diagnosis were significantly higher for acute otitis media, pharyngotonsillitis, pneumonia, SSI and UTI.

For the six most common diagnoses treated with antibiotics, a comparison of treatment choice per diagnosis with IH and OOH visits was made. The prescription rate was higher during OOH for pneumonia, acute otitis media, and pharyngotonsillitis. Although the difference was statistically significant, the choices of treatment for each diagnosis were comparable between IH and OOH prescriptions (Table 6).

During the study period, the level of infection visits was constant, but the antibiotic prescription rate decreased. Fewer prescriptions in children and for RTIs were the main reasons for the reduction. During OOH, there was a reduction both in infection visits and in antibiotic prescribing. The antibiotic prescription rate was higher during OOH than during IH, and when adjusting for age, sex, and diagnosis the difference was significant but small. The choices of treatments were similar.

This study showed that women visited primary care for infections more often than men and also received antibiotic treatment more often than men. The same pattern has been seen in other studies from Denmark, the Netherlands, and the United Kingdom [10, 27, 28]. The sex difference in the incidence of lower UTI was an important reason.

Our data on visit rates per 1000 inhabitants per years for infections were similar to the Primary Care Record of Infections in Sweden (PRIS) database [15], which consists of data since 2007 on visits with an infectious diagnosis and all antibiotic prescriptions from voluntarily participating PHCCs on an annual basis. Antibiotic prescriptions are in most cases linked to diagnoses and also includes information about age, sex, and laboratory results. The database has a larger dataset than in this study covering PHCCs in other regions but lacks OOH data. In the PRIS database, the visit rates per 1000 persons per year for infections during IH were 457 (in 2008), 441 (in 2010), and 406 (in 2013).

The total antibiotic prescribing in primary care decreased by 27% in this study. However, in the PRIS database [15] the reduction of IH antibiotic prescribing was 36%, as the IH antibiotic prescription per 1000 persons per year decreased from 245 (in 2008) to 201 (in 2010) to 157 (in 2013). For the corresponding years in our study, the IH antibiotic prescriptions per 1000 inhabitants were 212, 217, and 186, respectively. It is possible that participation in the PRIS database could have triggered a more restrictive antibiotic prescribing behaviour compared to our real-life study. A Finnish study [29] reported a 47% reduction in antibiotic prescriptions to children in primary and other out-patient care between 2010 and 2016, whereas our present study showed a 38% reduction in children in primary care between 2010 and 2014.

Several explanations are possible for the reduction in antibiotics prescriptions. For example, there might be increasing awareness among the general public that the use of antibiotics should be avoided when they are not needed. Also, physicians might have become more restrictive in prescribing. Another reason might be due to the antibiotic stewardship work performed by the Strama group, the Swedish strategic programme against antibiotic resistance [30]. In 2005, Strama together with the government launched a national strategy to prevent antibiotic resistance and healthcare-associated infections. Several actions have been performed in relation to this strategy. Diagnosis-specific guidelines for optimal antibiotic use have been published and promoted, and the use of antibiotics has been reported at the local, regional, and national level [17, 31]. During 2011–2014, the Swedish government ran a patient safety campaign aiming to decrease antibiotic use with the goal of fewer than 250 annual prescriptions in out-patient care per 1000 inhabitants for all prescribers together (primary and secondary care, dental care) resulting in a decrease from 385 prescriptions (2011) to 328 prescriptions (2014) [26, 32]. Furthermore, a pneumococcal conjugate vaccine was introduced in the Swedish national vaccination programme for children in 2009. Finally, a national economic bonus system was introduced for regions achieving a reduction in the antibiotic prescription levels, and incentive for quality outcome with the same goal was introduced in 2011 at the PHCC level in Kronoberg County.

During the period studied here, the number of OOH infection visits decreased by a third. Factors contributing to the decrease were shorter opening hours at the end of the study, a penalty fee (100 euros) introduced in 2008 for the PHCC for each patient attending the OOHC, and the introduction of a nurse triage system for walk-in patients at OOHC1.

The OOH antibiotic prescription rate per 1000 inhabitants per year was at the same level in the Netherlands, Sweden, and England (20, 28, and 31 prescriptions, respectively), but higher in Denmark (80 prescriptions) [9, 13, 27]. Two English studies have shown stable or increased OOH antibiotic prescription rates from 2010 to 2014 [8, 9]. In contrast, our study showed a decrease in antibiotic prescription rates.

The main explanation for excess prescribing during OOH is that infections that are often treated with antibiotics were more common during OOH visits such as acute media otitis, pharyngotonsillitis, and lower UTIs. The relative risk of antibiotic prescribing was decreased when adjusting for diagnoses. For SSI, the relative risk of receiving antibiotics during OOH remained elevated 1.20 (95% CI 1.18–1.23). It was uncommon to prescribe UTI antibiotics without a visit with infection diagnosis during OOH service (9% of UTI antibiotic prescriptions were without a visit during OOH compared to 39% during IH) although it was in line with current guidelines. This fully explained the higher UTI visit rate during OOH.

These results are similar to other European studies when comparing OOH and IH. A Norwegian comparison of tonsillitis and acute media otitis showed no difference in the prescription rate at OOHCs [33], and a Dutch study showed higher prescription levels during OOH for common infections and argued that the patients were sicker in the sense that they had more urgent problems that could not wait until the next day based on a revision of the EMR [13].

The remaining excess prescriptions during OOH after adjusting for diagnosis were estimated, leading to 2.2 more prescriptions per 1000 inhabitants per year compared to IH, which corresponds to 7.9% of the prescriptions during OOH and to 1.2% of all prescriptions during IH and OOH together. These prescriptions could partly be explained by sicker patients in need for urgent evaluation and an absence of control visits in the OOH setting. On the other hand, a reason could be a lower threshold to prescribe during OOH for example due to high workload or due to limited possibility to arrange for follow-ups.

Apart from the high relative risk of receiving antibiotics for SSI during OOH, there are no apparent areas to intervene. But because the total decrease of antibiotic prescriptions during the study period is 27% and the excess prescriptions during OOH are just above 1% of all antibiotic prescriptions, there would be limited gain from intervening in the OOH setting.

There were no differences in treatment choice, which corresponds with other quantitative studies from Norway and the Netherlands [13, 33] and with a Belgian qualitative study where physicians reported the treatment choice to be the same as during IH, although the threshold to prescribe was lower at OOHCs [11]. In contrast, an English study noted a higher proportion of broad-spectrum antibiotics during OOH [8].

Although the infection visit rate was unchanged, there was a significant reduction in antibiotic prescribing, especially to children and for RTIs. The increased antibiotic prescribing rate during OOH was small when adjusted for age, sex, and diagnosis, and no excess prescribing of broad-spectrum antibiotics was seen. Therefore, interventions selectively aiming at OOHCs seem to be unmotivated in a low-prescribing context.