Setting and population
This observational study compared patients residing in the Skåne region, Sweden’s third largest county with 1.4 million inhabitants. Patients were managed using “digi-physical” care or using traditional office visits at 16 primary health care centers across Skåne. Apart from the previously mentioned digital primary care providers, patients have the option to seek physical care at their primary health care center, which is usually open between 8 a.m. and 5 p.m. Patients can also seek care at out of hours clinics, open from 5 p.m. to 9 p.m., or visit the emergency department of any hospital. All index visits in the current study were conducted at Capio, one of Sweden’s largest primary health care center providers, which has adopted the “digi-physical” model since May 2017, using an eVisit platform developed by Doctrin AB. At the time of the study, Capio was the only known primary health care provider that offered both office visits and eVisits, while other eVisit providers simply referred patients who were deemed to require a physical examination. This meant the patient and physician had to restart the consultation, which resulted in two payments.
Inclusion criteria were visits with a chief complaint of sore throat, cough, cold/flu-symptoms or urinary symptoms as specified by free-form text, or visits with a documented International Classification of Disease code J030 (streptococcal tonsillitis), J069 (acute upper respiratory infection), or N300 (cystitis) [15]. Index visits were selected by identifying each patient’s earliest dated physician visit (for the chief complaints included) between March 30th, 2016 and March 29th, 2017 (office visits only) or between March 30th, 2018 and March 29th, 2019 (eVisits and office visits), i.e. before the COVID-19 pandemic. Exclusion criteria were patients aged < 18 years, residence outside of Skåne county, male patients with urinary symptoms and identifiable visits for similar chief complaints in the past 21 days. In addition to this, each patient was only allowed to contribute with one index-visit across the entire cohort. The earliest dated visit was chosen as the included index visit.
The platform
The eVisit platform assessed in this study can be accessed by patients through their smartphone, computer, or tablet seven days a week from 7 a.m. to 10 p.m. Patients choose their chief complaint and proceed to answer a set of symptom-specific questions. Answers are structured in a report presented to a physician who then initiates a two-way text-based communication within 15 min for medical decision-making, including staying available for observation (watchful awaiting) or utilizing “digi-physical” care by scheduling a physician appointment at a physical Capio primary health care center within 48 h if needed. The receiving physician at the primary health care center gets access to the same medical history generated by the eVisit platform and the text from the chat communication between the physician and the patient for an improved transition. Capio has protocols for each chief complaint, with indications for scheduling physical care and key performance indicators to follow-up protocol adherence.
Power calculation and recruitment
Previous research on office visits for upper respiratory tract symptoms reported a 26% two-week follow-up rate [16]. Using a binary outcome power calculation with a non-inferiority limit of 6.5%, an alpha level of 0.05, for 80% power, we estimated needing 564 visits per group. Informed consent was acquired from all included participants. eVisit patients were invited once and consented digitally prior to their visit. For office visit patients, data extraction software (by Medrave Software AB) was used to identify adult patients with key words in the electronic medical records free-form text corresponding to included chief complaints (Additional file 1). A random selection of identified office visit patients were invited through letters, including two reminders to non-responders, posted to their home address after their visit with a signed response returned in a prepaid envelope as previously described [15]. After acquired consent, remaining exclusion criteria were applied resulting in the final cohort (Fig. 1).
Data collection
Baseline data including chief complaint, visit date, age, sex, and patient residence were acquired from the medical record of the healthcare provider using the same data extraction software that identified patients. Automatically extracted data on chief complaints had previously been manually validated by reading all free-form text in the electronic medical record of the index visit for a subset of visits (n = 783) [15]. For eVisits only, data were also extracted regarding recommended follow-up by the physician as either self-care, continued eVisit, or recommended outpatient physical visit (urgent or non-urgent) as this was documented as part of the eVisit electronic medical record template. Patient data related to county-wide health care contacts within two weeks of their index visit were acquired from a county-wide registry (Region Skånes Vårddatabas, RSVD) registering all health care contacts billed to the local county council, including set diagnoses and health unit names for each health care contact. The database does not include visits provided through health care providers without a reimbursement contract with the local county council, but such visits only account for around 1% of all healthcare expenditure in Sweden [1].
The primary outcome was proportion of patients with one or more physical visits to a physician within two weeks after the first 48 h of their index visit, as “digi-physical care” per definition involves a proportion of visits inevitably proceeding to physical examination within 48 h of their eVisit assessment. Visits beyond 48 h after index thus represent visits not expected in the “digi-physical” model. To make subsequent utilization beyond this window was comparable to office visits, we excluded physical visits within 48 h of the index visit after both eVisits and office visits in the primary outcome. As most patient-initiated primary care contacts in Swedish primary care are initially managed through nurse telephone triage, the number of outpatient telephone contacts with nurses within two weeks of the index visit (not including the day of the index visit) was evaluated as a secondary outcome. Other secondary outcomes included proportion of additional outpatient physical visits within 48 h of index visit, visit location (primary care, out-of-hours visit, emergency department, or other outpatient clinic), and proportion of patients admitted for inpatient care.
For eVisit patients only, we also calculated secondary outcomes regarding proportions of index visits in which the patient was recommended self-care, continued digital care or physical follow-up, respectively. eVisit-physician-documented recommendation for an urgent visit within 48 h, a non-urgent primary care visit, and referral to other healthcare providers (including emergency departments) were all considered a physician recommendation for physical follow-up. In 13 cases where data regarding recommended follow-up were missing, data were manually collected through review of electronic medical records.
Statistical analyses
Analysis was conducted in IBM SPSS version 26. Visits with a chief complaint of sore throat, cough, and common cold/influenza were all grouped together to a “respiratory” group, while visits for urinary symptoms were considered a separate group.
Student’s t-tests were used to compare continuous data and were presented with mean and standard deviation. Chi-square test was used to compare categorical data, presented with percentage.
We hypothesized that there was no clinically relevant difference in the number of physical visits within two weeks when comparing eVisit patients to office visit patients, excluding the first 48 h where a larger portion of eVisits patients are expected to be encouraged to proceed to a physical visit. Hypothesis testing was conducted by comparing patients with index eVisits and index office visits, after excluding patients with subsequent physical visits within 48 h.
Sensitivity analyses were conducted comparing subsequent physical visits including visits within 48 h, but instead excluding eVisit patients recommended various levels of physical follow-up to evaluate robustness of findings.
As chief complaint and age may confound risk of further follow-up, multiple binary logistic regressions were conducted with physical visit or nurse phone contact as the dependent variable and visit type as the independent variable. Office visits were used as the reference group, with the enter regression models adjusted for age and chief complaint.
Exploratory subgroup analyses were conducted to evaluate health care utilization of eVisit patients who received various follow-up recommendations.
Further subgroup analyses were conducted to calculate the proportion of physical visits within various levels of care (ranked from highest to lowest acuity: emergency care, out of hours care (including ambulatory care), primary care, and other outpatient care) during the follow-up period. For patients in contact with multiple levels of care, the highest level of care was included.
Physical visit locations classified as emergency or other outpatient care were manually reviewed by looking up health unit names of the health care contacts as specified in RSVD to make sure the visit location was validly classified. For both groups, inpatient care within the entire follow-up period was also compared.
For a subset of patients with physical visits within two weeks (836 respiratory and 434 urinary complaints), the first three diagnoses recorded in the electronic medical record were manually reviewed together with a specialist in family medicine (SC and PM) and used to assess whether the visit was likely related to or unrelated to the index-visit.
The study was registered at clinicaltrials.gov (Identifier: NCT03474887) and reported using the STROBE-checklist.