All Swedish residents have a personal identification number, which includes their date of birth and gender, and this is used for identification in all official registers in the country. Moreover, all residents are covered by the National Health Insurance Act, which allows them to see any physician they choose at heavily reduced fees [19].

Drugs may be prescribed only by licensed physicians and by midwives and nurses who meet specific requirements [20]. All medical care in Sweden is organized within regions, geographically corresponding to counties. PHC (public or private), including physical-PHC or internet-PHC, is usually the first point of call for patients because of its accessibility [21].

Internet-PHCs were established in Sweden in 2016, including services allowing patients to access health care using a smartphone or computer [4]. The 4 largest digital health care companies in Region Sörmland have approximately 80% of the collective digital market share in the region. The Swedish digital care system is designed to work as part of the Swedish welfare system, which means that the region reimburses the main costs for health care, prescribed drugs, and also for internet-PHC.

The study was carried out in accordance with the tenets of the Declaration of Helsinki and the Swedish Ethical Review Authority (diary number 2020-04098) approved the study protocol. The previous local ethics committees have been closed down and since the 1990s have been replaced by the Ethical Review Authority, a government agent checking that ethics applications follow the law of research ethics. Data were collected in standard procedures in the Region Sörmland Administrative Office for financial reimbursement to PHC caregivers and quality assurance of PHC provided within the region. The approved study protocol allowed for analyses of these collected data.

Data for all appointments with physical-PHC and the 4 largest internet-PHCs in Region Sörmland from January 1, 2020, until March 31, 2021, irrespective of residency in Sweden, were obtained from the Region Sörmland Administrative Office. Data were collected in standard reports from the caregivers operating in the region. The data set included 3,064,680 appointment records for 1,238,183 patients, including 1,482,931 appointment records for 785,583 Sörmland and non-Sörmland residents consulting physical-PHC or the 4 largest internet-PHCs. Out of this material, the main study population for this report was 160,238 appointments for 68,332 Sörmland residents consulting regional physical-PHC or the 4 largest internet-PHCs. Information on the size of the Region Sörmland population by age and gender was obtained from the Swedish government agency Statistics Sweden. Data were analyzed using the Statistical Analysis System (SAS; version 9.3) [22].

Variables obtained for each appointment included the patient’s personal identification number, which provided information on gender and attained age, residential area code, care provider (physical-PHC or internet-PHC), appointment date, diagnosis according to the ICD-10 (International Statistical Classification of Diseases, Tenth Revision) [23], prescriptions according to Anatomical Therapeutic Chemical code [24], and date with respect to the COVID-19 pandemic. The data set was complete with no missing data.

Simple differences between continuous variables were tested with Student t test and differences between categorical variables with chi-square tests. The effects of care provider (physical-PHC or internet-PHC) on antibiotic treatments for 6 diagnoses according to Strama [25], adjusted for the effects of attained age at the time of appointment, gender, and pandemic period, were analyzed by descending logistic regression analysis (probability modeled as the presence of antibiotic treatment). The time in relation to the pandemic period (before or during) was defined as 0 (before March 1, 2020) and 1 (from March 1, 2020, onwards).

Statistical power was computed using the SAS power procedure. The statistical power of the study was >99.9%. The SAS logistic regression procedure provided information on estimates and their 95% CIs, Wald chi-square (test parameter), odds ratios and their 95% CIs, the concordance index C (a measure of explanation similar to the receiver operating characteristic curve), and P values. A large number of statistical tests were performed, and Bonferroni adjustment was made to avoid mass significance issues. P<.005 was therefore used to indicate statistical significance.

The results of this study showed that the patients differ in age and gender between internet-PHC and physical-PHC. Infection diagnoses were more common in internet-PHC (22.3%, 7979/35,840 of appointments) than physical-PHC (8.4%, 10,454/124,398) but antibiotics were prescribed in less degree for infection diagnoses appointments, internet-PHC, 25.8% (2057/7979) versus 61.3% (6412/10,454) for physical-PHC appointments. On average, internet-PHC providers prescribed antibiotics for 6 infection diagnoses 70% less often than physical-PHC providers (OR 0.23-0.39), after taking account of the effects of differences in gender and age distributions between providers, and the period in relation to the COVID-19 pandemic. The results were similar for all 6 diagnoses tested in multivariate analysis.

The strengths of this study included the use of data that were representative of a defined geographical area, thus minimizing the risk of selection bias regarding the included population. Moreover, the study population was large enough to avoid problems associated with low statistical power. Because the outcome variables were dichotomous, logistic regression was used to obtain the main results. The study limitations included the lack of access to medical record data. However, there was no reason to believe that the registration of diagnoses and treatments would be biased to an extent that would affect the results.

Prescribing of antibiotics in Sweden decreased by more than a quarter over the past 10 years [26], and strategies and goals to decrease antibiotic use are also set globally to decrease antimicrobial resistance.

These results indicated that infection diagnoses were much more frequent in internet-PHC appointments (22.3% [7979/35,840] of all appointments) compared with physical-PHC (8.4% [10,454/124,398]). This might be because these diagnoses are more common among younger people. Most patients chose internet-PHC for acne, impetigo, acute cystitis, and upper respiratory diseases. This might be an age effect, but these are also diagnoses that can be made by medical history or video observations. The high number of upper respiratory disease diagnoses may also be related to the COVID-19 pandemic.

Considering the effect of the pandemic, the degree of infection diagnoses by physical-PHC dropped from 12.7% to 7.6%, compared with 25.5% to 22% for internet-PHC. Hygiene recommendations probably decreased the transmission of common infections and physical appointments [27], with patients with an infection diagnosis shifting from physical-PHC before the pandemic (2629 physical- vs 848 internet-based) to internet-PHC during the pandemic (7825 physical vs 7131 internet appointments).

This study demonstrated that internet-PHC providers prescribed fewer antibiotics than physical-PHC providers during the study period, in accordance with earlier, smaller studies in Sweden [12], but in contrast to reports in the Swedish daily press [10]. The higher and broader pattern of antibiotic prescriptions in physical-PHC is probably related to the broader range of diagnoses and possibly to laboratory test results and more complicated or multiple diseases [25].

Antibiotic prescribing levels for defined infectious diseases were significantly lower for internet-PHC for all 17 Strama [25] defined infections studied. Various respiratory infections were rarely treated with antibiotics by internet-PHC, following Strama’s recommendations (<5%), despite the fact that the antibiotic prescription limit is not always possible or likely to be met [25].

Infection diagnoses often need to be confirmed by laboratory tests before prescribing, leading to less prescribing in internet-PHC. Diagnoses for which antibiotics were frequently prescribed included acute cystitis (82.1% of diagnosed) and Lyme borreliosis (85.5%), which was in line with Strama recommendations [25] for internet-PHCs. Patients with these diagnoses were prescribed antibiotics more frequently in physical-PHC.

Uncomplicated acute cystitis can be diagnosed only with an online appointment and treated with the first-line antibiotics nitrofurantoin and pivmecillinam, without laboratory testing, using internet-PHC. This appeared to be the case for other infections as well as with a higher degree of prescribing for visually diagnosable infections, such as skin infections and acne. Patients may also initiate contact with internet-PHC providers to change the administration formula after earlier contact with a physical-PHC provider. Further studies are therefore needed to establish the timing for those who make the first diagnosis and when antibiotics are prescribed to determine if the recommended Strama <5% level is adequate.

Internet appointments have several advantages, of which the most important is quick access. Internet appointments save time for the patients, are comfortable, suitable, and seem to be sufficient for making simpler diagnoses and issuing prescriptions. Internet-PHC appointments increased during the COVID-19 pandemic to avoid exposure to other patients in waiting rooms or in-person consultations. In Sweden, prescribing is almost totally digitalized, making internet-PHC prescribing easier, with patients receiving their prescriptions online instantly.

In addition to the above advantages of internet-PHC, it may have disadvantages, including leading to more appointments for symptoms that would not lead to a physical-PHC appointment. Younger patients and women tend to use internet-PHC more often due to a higher degree of digital literacy, whereas the technical threshold may exclude elderly and disabled patients. Further studies should determine if an increase in internet-PHC use releases physical-PHC resources for these groups. Internet-based PHC may fill a gap in the health care system and physical-PHC and internet-PHC may complement each other.

This study demonstrated substantial differences in the characteristics of patients who consult internet-PHC and physical-PHC providers in Sweden. The proportion of patients who chose internet-PHC increased during the COVID-19 pandemic, but the rate of infection diagnoses decreased in internet-PHC. The probability of a patient with an infection diagnosis receiving antibiotics after an internet-PHC visit was reduced compared with physical-PHC, after adjusting for attained age at the time of appointment, gender, and the pandemic. Internet physicians thus prescribe fewer antibiotics than physicians in physical consultations and show appropriate prescribing according to the recommendations.