Modelling within-individual variability of clinical parameters and investigation of their effects on clinical decision-making

All measured clinical parameters vary within individual the same individual. This variation includes biological variation (chance variation around a physiological mean) and variation in the measurement process (measurement error). Within-individual variation is typically measured as the coefficient of variation (CV_i) which is the standard deviation of within-individual measurements divided by the mean of within-individual measurements.

Implications of within-individual variation

Within-individual variation has important implications for clinical decision-making. If a diagnosis is based on a measurement, chance variation means there is a probability of misdiagnosis. If clinical decisions when managing a long-term condition are informed by ongoing measurements during patient follow up, chance variation means there is a probability the clinician will make incorrect inferences about the condition. For example mistaking chance variation in measurement as a response to treatment, or a failure to respond to treatment, deterioration in the condition or that the condition has not changed. The effects of chance variation on clinical decision-making have not been quantified.

Recent research

Recent research by a doctoral student has reviewed previously published studies of the variability of a number of clinical parameters and then quantified the observed variability of these clinical parameters in a large dataset of UK electronic primary care records. The research focused a selection of on clinical parameters which are measured in different ways. Two are laboratory blood tests, HbA1c (glycosylated haemoglobin), a measure of glycaemic control used for the diagnosis and monitoring of diabetes, and CRP (C-reactive protein), a measure of inflammation is used to distinguish between acute respiratory infections caused by viruses and bacteria and to monitor long-term inflammatory conditions. One is clinical measurement of respiratory function: FEV1 (forced expiratory volume) and FVC (forced vital capacity). The others are patient reported questionnaires to measure depression symptomatology: the PHQ-9 (Patient Health Questionnaire-9), Becks Depression Inventory (BDI) and the HADS (Hospital Anxiety and Depression Scale).

The observed within-individual variation of all of the clinical parameters investigated is greater in electronic primary care records than is appreciated from previous published data (small studies with limited follow up). Within-individual variation of all of the clinical parameters investigated also differed between different subgroups of patients: between those with and without a related diagnosis, those with different sociodemographic characteristics. Variability in the long-term was also affected by long-term (age-related) trends in the underlying clinical parameter. This substantially complicates the understanding of expected within-individual variability in different types of patients and complicates the estimation of potential effects of within-individual variability on clinical decision-making.

Aim of this doctoral thesis

This doctoral thesis will undertake statistical modelling of a large dataset of electronic primary care records to understand the factors affecting within-individual variation in measured clinical parameters. It will focus on the previously identified parameters:

·      HbA1c (glycosylated haemoglobin)

·      CRP (C-reactive protein)

·      FEV1 (forced expiratory volume) and FVC (forced vital capacity)

·      PHQ-9 (Patient Health Questionnaire-9), BDI (Becks Depression Inventory) and HADS (Hospital Anxiety and Depression Scale)

The thesis will then investigate the likely effects of within-individual variability of these clinical parameters on clinical decision making. This will make use of statistical modelling and online clinical vignettes where clinicians are asked to make clinical decisions which are informed by clinical-parameters showing an empirically accurate degree of within-individual variation.