The circumstances of the pandemic may have inspired new trends in healthcare that will remain in the long term, such as telehealth adoption, while others will more likely return to the pre-pandemic norm. These behavioural changes, both short and long term, should be considered when observing population health trends and drawing conclusions in the future for real-world studies.

Telehealth options had been available in limited instances prior to the pandemic, but the sudden demand for the alternative, socially distanced approach to care spurred a rapid increase of virtual-only visits. According to a CDC report, the number of telehealth visits increased by 50% during the first quarter of 2020, with a 154% increase in visits noted in the last week of March, compared with the same period in 2019. The two mediums differ both in the treated population and the care that is administered, meaning some patients may have opted out of virtual visits because they were more comfortable in an office setting, and some doctors may have held off on diagnosing patients virtually when blood work, physical exams and vitals were needed to accurately determine treatments.

Patients changed their prescription approaches, which resulted in a general decrease in prescription refills overall. These changes included extended medication fills to reduce the number of times they would need a refill or transitioning to in-home or self-administered forms of infusion therapies.

Pandemic-related job loss prompted a significant shift in insurance coverage, with over 23 million Americans becoming unemployed by April 2020, which could impact longitudinal data continuity. The impact is due not only to shifts in those seeking or receiving care due to the loss of insurance coverage, but also the way data is captured through medical claims.

Patients in under-represented and vulnerable populations such as those who are older, disabled, located in rural areas or in lower socioeconomic groups may have been at a disadvantage in this shift to more virtual care due to access issues. That said, many of these populations also struggle to receive in-person healthcare due to lack of transportation and insurance coverage, which should remain a consideration post-pandemic.

Whether it’s the rise of mental health conditions, obesity, drug dependency, sedentary behaviours or increased stress, social isolation may have impacted the severity of disease epidemiology and adverse clinical outcomes. Conversely, precautionary activities such as wearing masks, sanitising hands and surfaces, and avoiding close contact have also decreased the incidence of influenza, sexually transmitted diseases and respiratory syncytial virus (RSV).

Disease prevalence errors:
Limited access to care during the pandemic interrupted routine care, lab work and medical screenings, which decreased overall volume of medical interactions. By mid-April 2020, medical claims for routine cancer screenings dropped by 39% to 90% compared to before the pandemic. Data from this period may spur an initial perceived decrease in disease prevalence and medical utilisation. However, there will likely be an increase as regular care is resumed.

Misrepresented disease severity:
Those who had delayed cancer screenings or lab work have a higher likelihood of a later stage prognosis. This will impact the kind of treatments available to them and their response to those treatments. If context of the pandemic is not taken into consideration when analysing data, it will incorrectly reflect an overall increase in disease severity.

Incorrect attribution of prescription delays:
It is important to understand the root of the delay, whether the patient missed refills due to fear of COVID-19, there was reduced availability of certain products through pharmacies, or a hesitancy by physicians to administer new therapies without an in-person examination. If patients experienced disruptions in their medication regimen, it has the potential to exacerbate their conditions or force them to misuse over-the-counter medications.

Changing processes to account for the underlying catalysts of pandemic-era behaviour shifts will be important to generate reliable RWE, as well as create a lasting impact on the acceptance of RWE in the years to come. Researchers should consider the steps to ensure the RWE they generate is reflective of the circumstances of the time.

Point in time analysis:
Describe the population of interest at different points in time, noting those that correspond to the pandemic. This can help put results in the context of those who are seeking care.

Restricting parameters:
Take the time period and type of care received into consideration, restricting the data to a certain type of care. Alternatively, the data collected during the pandemic could be handled separately from previous and subsequent years to address the effects of the pandemic on outcomes ascertainment.

Address potential biases:
Account for potential biases among a population due to the differential impact of the pandemic on certain subpopulations. This will help researchers validate the insights generated during the pandemic.

Expand characteristics:
Consider the inclusion of added characteristics of the population into the multivariate analyses (ie, care received, diagnosis or evidence of prior or current COVID-19 infection, geographic location of patients and calendar time), as well as other variables affected by the pandemic and COVID-19 infection, which will lead to more comprehensive analyses.

COVID-19 has had far-reaching impacts for all healthcare users, from patients wanting to receive routine care such as vaccines and patients in need of mental health treatments, to cancer patients receiving acute care. While this guidance is not exhaustive, broaching these considerations can help spur critical thinking within the real-world research community so that they can continue to conduct research with a high level of confidence that their findings accurately represent the patterns from the pandemic era.