Table 4.

Some limitations of electronic health records for COVID-19 research, and how they may impact on the relative risk and incidence of neuropsychiatric disorders

Limitation	Effect on relative risk of NPD	Effect on incidence of NPD after COVID-19	Comments and implications
Undocumented COVID-19 infection (in comparator cohort)	Decreased	Unaffected	Some subjects in comparator cohorts will have had COVID-19
People with asymptomatic COVID-19 under-represented	Unaltered if comparator cohort has a similar limitation (e.g. other respiratory infection). Increased if comparator cohort does not have a similar limitation (e.g. general population or all people testing negative for COVID-19) and if NPDs are less common after asymptomatic COVID-19	Increased, if NPDs are less common after asymptomatic COVID-19	Using multiple comparator cohorts can help better characterize relative risk. Prospective studies specifically targeting asymptomatic COVID-19 cases can help determine the incidence of NPDs in that subgroup.
NPD might have started before the index event but not diagnosed till afterwards	Unlikely	Increased	Could lead to misattribution of causality to COVID-19 or other index event
NPD present but not yet diagnosed (e.g. because of diagnostic delay or because affected individual does not seek medical attention)	Unlikely	Decreased	The effect on incidence will tend to counter the over-estimation arising from the limitation above. Active follow-up in prospective studies is required to detect these cases
Incomplete past health records	Unlikely	First-onset diagnostic rates increased if previous episodes had occurred but were not recorded	Less of a problem in countries where people have a single health record
Lack of information on viral strain	Could be increased, decreased or unchanged	Could be increased, decreased or unchanged	Effects depend on whether the strain affects risk of NPD and, if so, the timing and design of the study
SARS-CoV-2 could affect patient or physician behaviour and thence diagnostic behaviour	Could be increased, decreased or unchanged	Could be increased, decreased or unchanged	Direction of effect depends on what (if anything) COVID-19 does to health-seeking or physician behaviour. Difficult limitation to overcome.
People making no further contact with health services considered lost to follow-up	Could be increased, decreased, or unchanged	Decreased when using Kaplan–Meier estimator, unchanged when using total counts	Prospective studies with active follow-up and low attrition can address this limitation
Unknown NPD severity, course, and outcome	None	None	Difficult to assess using conventional EHR data. Although risk and incidence unaffected, these factors have implications for treatment and services

NPD = neuropsychiatric disorder.