Protocol for the Development and Analysis of the Oxford and Reading Cognitive Comorbidity, Frailty and Ageing Research Database-Electronic Patient Records (ORCHARD-EPR)

The advent of hospital electronic patient records (EPRs) offers the opportunity to exploit large-scale routinely acquired clinical data to improve patient care. Such data offer several advantages. First, they can be obtained at scale without additional burden to patients or staff and at relatively low cost. Second, the use of EPR data (with the appropriate governance and ethical approvals) allows the entire hospital population to be studied without the unavoidable selection bias incurred in requiring individual consent for research assessments.1 This is particularly advantageous in studies on older, frail and cognitively impaired patients in whom capacity to consent to research may be lacking, or ability to tolerate research assessments may be limited.2 3 Third, obtaining data from the EPRs provides a considerably richer source of data than is currently available from hospital administrative data sets that are comprised largely of ICD-10-coded diagnoses, which have limitations especially in capturing complexity including frailty syndromes.2 4 5 Fourth, EPRs provide real-time information allowing implementation of algorithms to identify those at high risk (eg, of delirium).6

The case-mix of general hospitals is changing across high-income countries in line with population ageing with implications for a wide range of hospital specialties.7–9 Older people aged ≥65 years occupy the majority of hospital bed-days and many have complex conditions.10 Up to a half have a cognitive disorder (eg, delirium, dementia) and around 40% are physically frail with an increased vulnerability to stressors such as a sudden decline in health.2 11–16 Both cognitive and physical frailty are associated with a broad range of adverse health and social outcomes2 11 13 15 16 and improving acute hospital care for older patients is a priority as reflected in policy documents and guidelines internationally.17–22 Therefore, hospital data systems need to be established to identify frail patients across the hospital and across specialties, discriminate cognitive from physical frailty and ideally, the frailty domains affected given the implications for care.2 11 17–22 In addition, collection of detailed clinical data including on diagnosis, comorbidities and illness acuity will aid understanding of the prognostic value of frailty across a range of acute hospital specialties and settings, improve risk stratification, inform secondary and tertiary prevention measures and provide data for research.2 11 20

In this protocol, we describe the methods underpinning the Oxford and Reading Cognitive Comorbidity, Frailty and Ageing Research Database—Electronic Patient Records (ORCHARD-EPR), which uses routinely acquired rich multimodal electronic individual patient data including routine cognitive/delirium screening, nursing risk assessments (falls risk, nutrition, pressure sore risk), observations, laboratory tests and brain imaging. ORCHARD-EPR will facilitate healthcare innovation and medical decision-making including the development/validation of risk prediction algorithms including for delirium and future dementia. ORCHARD-EPR will also enable key evidence gaps to be addressed around cognitive and physical frailty in the hospital setting as summarised in box 1 overcoming the limitations in existing studies from small selected samples (prospective studies) or reliance on hospital diagnostic ICD-10 coding (large retrospective studies).4 5

Box 1

Knowledge gaps to be addressed using ORCHARD-EPR

Cognitive frailty

Prevalence of all-cause and subtypes of cognitive frailty in the acute hospital population overall and by age, sex and specialty.

Accuracy of ICD-10 coding for cognitive diagnoses.

Outcomes of all-cause cognitive frailty and by subtype (adjusted for confounding).

Overall frailty

Prevalence in the acute hospital population overall and by age, sex and specialty including same day emergency care (SDEC) as measured by different frailty tools (eg, Clinical Frailty Scale, Hospital Frailty Risk Score, Dr Foster score, frailty aggregate score).

Prevalence of frailty markers identified from nursing assessments (eg, at risk of falls, pressure sores and malnutrition).

Time trends in frailty prevalence in in-patient vs SDEC populations.

Outcomes of frailty (adjusted for comorbidity burden, illness severity, comorbid dementia, care home residence).

Prevalence and prognosis in key subgroups, that is, care home residents, comorbid dementia/delirium, severe illness.

Healthcare innovation

Development of predictive algorithms for delirium, and for dementia on follow-up.

Use of routinely acquired brain imaging for the development of AI tools to quantify ‘brain frailty’ including atrophy and small-vessel disease and old stroke lesions to aid dementia subtyping and risk stratification.

Real-time tracking of frail individuals through the hospital system to inform service delivery and design.

Development of a real-time, ‘streamlined Comprehensive Geriatric Assessment’ to identify care needs using routinely acquired electronic patient record data.

This post was originally published on