This blog was created to keep healthcare professionals, researchers, methodologists, and patients up to date with the latest primary care research. For more information about the Research Institute, visit our website;

Wednesday, 20 December 2017

Multi-state modelling: a brief introduction

Written by Chris Morton | PhD Student | | @PCSciences

Multi-state modelling, a flexible framework which describes complex clinical processes over time, is often overlooked by the more favourable survival and longitudinal models. In this blog, I hope to provide an accessible introduction to multi-state modelling, drawing on my recent dissertation for my MSc at Lancaster, whilst also introducing my current research. 

What is a multi-state model? 

The concepts of multi-state models may be more familiar to researchers than they realise. 

In a multi-state model, an individual/patient falls under one of several 'states', and may transition between those states over the course of their lifetime. 

So in a 3-state illness-death model, a healthy patient is in state 1, they transition to stage 2 if they become diseased and stage 3 reflects death. Traditional survival analysis can be considered a 2-state (alive or deceased) model. 

Health researchers are often interested though, in the effect of covariates (e.g. treatment regime) on the risk of an event at each instant of time. For a general multi-site model, the event is the transition from one state to another, and it is possible for separate baseline risk and covariate effects to be associated with each transition. 

Why use multi-site models in primary care? 

Multi-state models are commonly applied to clinical conditions where there is an increasing state of disease severity which precedes eventual death. Only using this approach though, it unnecessarily limiting - there are a wide range of potential applications of multi-site models in a primary care setting (e.g. mental health conditions and patient habits such as smoking and alcohol consumption). 

The biggest advantage of using a multi-site approach is the insight you can gain about every aspect of a process. For example, we can separately measure the factors affecting a smoker's decision to quit and those affecting a relapse into smoking. 

For my dissertation, the illness-death model is used to describe the progress of intact dental veneers (state 1), which may at first become discoloured (state 2) before eventually fracturing (state 3). My methodology focuses on the methodology rather than this particular application, but the example illustrates the diversity of potential uses for a multi-state approach. 

What are the key steps in model building? 

1) Representing clinical processes

The first step for a researcher is deciding how to represent a clinical process in terms of a discrete state and the permitted transitions between them - which isn't always straightforward. 

For example, when following the habits of a smoker, do two states ('current smoker' and 'current abstinence') adequately capture their behaviour? Or, do we want at least three - 'habitual smoker', 'withdrawal phase' and 'long-term abstinence' (see paper). 

2)  Matching covariates with state transitions

Consider which covariates act on each state transition and whether their effect should be constrained to be equal amongst transitions (due to a biological rationale, or to simplify model calculation). 

3) Model Fitting  

A natural starting point for model fitting is using freely available packages in R software: 'mstate' and 'msm'. Mstate' can be used if state transitions are observed at an exact date, such as when a patient dies. 'Msm' can handle panel data, such as when a patient is known only to have changed state at a time since their previous clinical observation. 

Beyond the functionality of the above packages, lies potential complications which were the focus of my dissertation. I analysed panel data with a clear dependence structure (multiple veneers in the same patient) and showing evidence of time-inhomogeneity (the instantaneous risk of a state transition changes over time). 

Unfortunately though, some of these more complicated models remain inaccessible to many researchers, largely due to a lack of software availability. 

What next? 

Fresh from a MSc in Statistics at Lancaster University, I now face the trials and tribulations of a PhD at the Research Institute for Primary Care and Health Sciences at Keele. My research investigates the treatment and symptom patterns of patients with polymyalgia rheumatica (PMR) and what factors predict these. 

Patients go from being treated with steroids to a state of remission, which may be followed by future relapses, again bringing to mind a multi-state framework. It is still too early to say whether such methods will provide me with an informative perspective, and research trajectories don't always run so smoothly. 

Regardless, multi-state modelling, together with longitudinal methods learned at Lancaster will always be a welcome tool to have in my future career as a primary care researcher, and I hope this blog has you already considering further applications for this interesting and useful area of methodology. 

About the author: 
Chris has recently started his SPCR funded PhD at Keele University, entitled 'early symptoms and treatment duration in polymyalgia rheumatica: a joint modelling approach'. In the previous academic year, Chris completed an MSc in Statistics at Lancaster University, for which he won two departmental prizes: the Tessella Industry Prize for 'Best Computational MSc Statistics Dissertation' and also a Postgraduate Statistics Centre Prize for 'Learning Excellence.'

Pain on a platform: STarT Back in Seattle

Written by Nicola Evans | Implementation Manager | | @PCSciences

Anyone familiar with the Research Institute will almost certainly have heard of the 'STarT Back' tool mentioned on more than one occassion. But for those unlucky few, the tool was established back in 2008 with the overarching aim to improve the care that patients suffering with back pain receive. Using a stratified (systematic) approach, the tool allows healthcare providers to suitably match the right treatment to patients according to their risk of experiencing persistent disabling low back pain. 

The STarT Back tool has certainly begun to put back pain on the healthcare agenda, and the tool is being implemented more and more by clinicians, commissioners and researchers all over the world. There is still a long way to go though, and through building collaborations we're able to address the currently challenges surrounding back pain. 

Myself and senior Knowledge Mobilisation Fellow, Kay Stevenson, recently spent two day ins Seattle to discuss the implementation of STarT Back with key collaborators, Kaiser Permanente. 

Although there is a longstanding history of collaboration between the Research institute and Kaiser Permanente (previously known as ‘Group Health’), most recently our focus has shifted to the world of implementation. Recent discussions shed light on a mutual concern with lack of materials available to explain chronic pain for patients, and a lack of training for primary care practitioners to give or support these explanations. 

The team at the Research Institute have tried to tackle this through the development of the STarT Back website, as well delivering training. Our collaboration with the Kaiser Permanente has led to a considerable impact of the STarT Back tool within the US and other countries, and recent discussions evolved into the development of a web-based ‘Pain Platform’ which will have a much wider reach and spread the word about the tool far and wide. 

The two-day visit enabled both teams to gain a better understanding of the relative perspectives and systems within chronic pain, and explore ways in which the ‘pain platform’ project will be taken forward.  

It was two wonderful days of sharing and learning, with both teams presenting their current implementation work and models for best practice in the area of chronic pain. Day two’s focus was future collaboration projects to strengthen implementation work that’s already been delivered. Our very own Professor Peter Croft introduced a seminar delivered by myself and Kay which highlighted the research-into-practice story - using STarT Back as an example of highlighting work within the NHS, finishing with a clear and strong story of the Impact Accelerator Unit and patient engagement. 

The visit was a clear success, enjoyed by both teams who built key strengths to help identify a plan to move forward. Clear opportunity for long-term collaboration on educational and training initiatives for clinicians involved in consultations and implementation research were also identified.

About the author: 
Nicki's role within the Impact Accelerator Unit is to support the operational delivery of key innovation projects on behalf of the Research Institute for Primary Care and Health Sciences (iPCHS). She takes the lead for the project management in the Unit, working closely with clinicians in the team as well as key groups within the NHS to ensure effective adoption and spread of our work. 

She graduated from University of Liverpool in 2001 with a degree in Nutrition and has worked in the NHS for over 10 years in the public health arena prior to joining the Impact Accelerator Unit in September 2015.  

Friday, 1 December 2017

'Are we there yet?' Saturation in qualitative research and how to use it

Written by Dr Ben Saunders | | @PCSciences

I think I speak for most qualitative researchers when I say saturation is taken for granted, yet there continues to be inconsistency in how it’s used and even uncertainty on how to use it. Drawing on our recent published paper, this blog tackles the ‘what, where, when, why and hows’ of using saturation in research.  

What is Saturation?

In short, saturation is used as a criterion for qualitative researchers to decipher when data collection or analysis is discontinued – in essence they reach ‘saturation point’. But this is research methodology we’re talking about, so it’s never going to be short and sweet.

From analysing existing literature, we’ve been able to identify four approaches to saturation, which may be either inductive (exploratory - finding patterns of explanations/theories) or deductive (using data to test pre-determined theories);

Theoretical saturation (inductive)

Data analysis leads to well-developed theories where no aspects remain hypothetical. The researcher(s) reaches the decision that no further data collection is needed.

Inductive thematic saturation

Similar to theoretical saturation, but it focusses on the identification and number of ‘new’ codes and themes rather than the completeness of theoretical categories.

A priori thematic saturation (deductive)

Establishing whether there is enough data to illustrate a pre-determined theoretical category.

Data saturation

Moving away from data analysis, this is based on how much data (i.e. the number of interviews) are needed until you’re no longer finding anything new.

Where and why should we use saturation?

The role of the theory is hugely influential to the relevance and meaning of saturation. In both deductive and inductive approaches, we are able to make sense of the role of saturation because of the underlying approach to the analysis being thematic. It will usually occur in interview or focus group studies that involve a number of informants.

It’s less straightforward in studies based on biographical or narrative approaches to analysis because they focus exclusively or predominantly on the accounts of an individual (e.g. interpretative phenomenological analysis). It might appear that saturation indicates completeness of a biographical account, but this is questionable. Can saturation usefully describe a participant’s story as ‘complete’ given the distance that it moves us away from the use of saturation in thematic approaches? Surely this would stretch the coherence and utility of saturation too widely? 

When and how?

Perspective will have a big implication on when saturation should be sought. Saturation can be identified early on if you take the ‘data saturation’ or/and ‘inductive thematic saturation’ approach. Data saturation will rely on the researcher’s perspective on what they’ve heard during interviews to decide whether further data collection is needed. Inductive thematic saturation looks for the (non)emergence of codes or themes.

In contrast, theoretical saturation is reached much later, often when grounded theory categories has been developed, so analysis is much more advanced.

Straus and Corbin quite rightly highlighted the issues with identifying the point of saturation and whether it’s just a cumulative judgement. Although it’s commonly seen as a discrete event, there will always be the potential for ‘new’ codes or theories to emerge. Analysis will therefore not suddenly become ‘rich’ or ‘insightful’ after the addition of one interview, rather it becomes ‘richer’ or ‘more insightful- raising the question ‘how much saturation is enough’ rather than ‘has saturation occurred’.

Attempting to identify the right point of saturation perhaps reflects the uncertainty of how to use it. Determining whether further data collection or analysis is needed, based on the data already gathered, essentially refers to the unobserved based on the observed – an uncertain predictive claim which can only be tested if the decision to halt data collection is overturned.

So, if saturation is unpredictable, what’s our advice? Here’s our top 4 points to consider when using saturation in your research….


Define its purpose

The relevance of your saturation and its meaning will depend on the role of the theory and the analytical approaches you’ve adopted etc. Therefore, it may serve different purposes for different types of research – purposes that need to be clearly articulated by the researcher.

Don’t saturate your saturation

There needs to be a limit on your range of purposes for saturation, or else you run the risk of stretching or diluting its meaning to the point where it becomes too widely encompassing.

Event or ongoing process?

When to use saturation, and how you reach saturation will differ depending on the type of study. Assumptions about whether it represents a distinct event or ongoing process will also differ.

Recognise its inconsistencies

If anything, this paper confirmed the need for a more transparent reporting of saturation, as well as a thorough re-evaluation of how it’s considered and used – including the recognition of potential inconsistencies and contradictions in its use. Considering the four different types of saturation outlined earlier will help serve as a guide for this.

The blog was adapted from the following paper:
Saunders, B., Sim, J., Kingstone, T., Baker, S., Waterfield, J., Bartlam, B., Burroughs, H. and Jinks, C. (2017). Saturation in qualitative research: exploring its conceptualization and operationalization. Quality and Quantity.

About the Author
Dr Ben Saunders is a Qualitative Research Associate at the Research Institute on the Stratified Primary Care Programme, and is an active member of the Social Science team. He is involved in the development and testing of a new stratified care intervention for treating common musculoskeletal conditions in primary care, and is also working on a number of research projects. Before joining the Research Institute, Ben completed his PhD at Cardiff University where he researched young adults’ experiences of living with long-term conditions, focusing on inflammatory bowel disease (IBD) and Type 1 diabetes. Ben currently supervises PhD students in the areas of stratified care, dementia caregiving, and young people’s experiences of stoma care.