Table 1.

Estimates of problem list completion rates in the existing literature.

Study	Setting	Condition(s) studied	Method for comparison (‘gold standard’)	EHR system used	Percentage of diagnoses included in problem list
Wang et al., 2020 [25]	383,404 patients in Partners Healthcare System, USA	Asthma, Crohn’s disease, depression, diabetes, epilepsy, hypertension, schizophrenia, ulcerative colitis	Algorithm based on ICD-10 diagnoses and prescriptions	Epic	72.9−93.5% (unweighted mean 85.5%)
Wright et al., 2015 [26]	160,341 patients in 10 healthcare organisations in USA, UK and Argentina	Diabetes	Haemoglobin A1c ⩾ 7.0%	Mixture of EHR Systems: Epic (n = 3), Allscripts (n = 2), EMIS (n = 1) and self-developed EHR systems (n = 5)	60.2−99.4% (weighted mean 78.2%)
Wright et al., 2011 [27]	100,000 patients at a single hospital in USA	17 medical conditions	Associations with medication and laboratory results	N/A	4.7−76.2% (unweighted mean 51.7%)
Polubriaginof et al., 2016 [15]	1472 patients in a single hospital in USA	59 medical conditions	Self-reported past medical history using a tablet questionnaire	N/A	54.2%
Hoffman et al., 2002 [28]	148 patients attending a general medicine clinic at a university affiliated Veterans Affairs hospital	9 diagnoses relevant to the choice of drug therapy for hypertension	Sensitivity and specificity of diagnoses recorded in the problem list of electronic records, with medical charts as the standard for comparison	VISTA (Veterans Health Information Systems and Technology Architecture)	42–81% (unweighted mean 62.4%)
Current study	516 inpatients with COVID-19 in a London teaching hospital	Any medical condition relevant to ongoing care	Manual review of free text medical records	Epic	62.3%