Over the past ten years, I have led a small development team at University Hospital Southampton NHS Foundation Trust in the creation and implementation of a radical approach to the Electronic Patient Record interface. Our product, UHS Lifelines [1], is a prime exemplar of the Stacked, Synchronised Timeline and Iconographic (SSTI) structure. It is live and in daily use across the hospital for some 2.5M individual patient record sets.

This system has evolved from the Lifelines project work of the Human Computer Interaction Laboratory of the University of Maryland in the mid 1990s [2], but which was not further developed into practical applications at that time.

The UHS Lifelines system permits the display, navigation and interrogation of all electronic records (documents, reports and events) for any one patient on a single computer screen, where the X axis displays continuously incremental time, and the Y axis displays the subject taxonomy for which e-data is available for the individual patient. Each icon acts as a dynamic window to the underlying document or cluster of documents, reports and events.

The UHS Lifelines sits efficiently over the data sets, which may be sourced from a number of different packages and software systems, so as to load the interface in real time.

The image is a live screenshot in an outpatient clinic of an older lady who has multiple co-morbidities and who has needed multiple hospital admissions. Each icon opens a frame to display the underlying content

The purpose of dynamic data visualisation

Data exists to help human beings to make decisions in the real world. A well designed data visualisation system will minimise the time and effort needed to reconstruct the story from the data, while maximising the quality of the decision from an overview of the entire data set. This fits the mantra of Emeritus Professor Ben Shneiderman of the University of Maryland, who mandated that any data visualisation system should start with a global Overview of the data, while allowing the observer to Zoom In on any point of interest, Filter Out extraneous material, and secure Details on Demand [3]. 

The core purpose of UHS Lifelines is therefore clear. It is to help the clinician to make better clinical decisions in less time, at least risk and with less fatigue, than would be possible with any other format of electronic information retrieval and display. An excellent data visualisation system efficiently conveys primary information through visual imagery in the form of patterns, colours, shapes and relative positions. This format directly addresses the fast processing powers of the visual cortex in the brain, before a word is read.

Paradoxically, the format of UHS Lifelines also (in effect) abolishes time. A document of any age can be located and opened at the same speed as a contemporary document, without the need to search through back catalogues, or across multiple screens, windows, tabs and lists, to find historic records.

The wider potential of the SSTI (Lifelines) interface

It will be immediately apparent that UHS Lifelines is the first-in-class exemplar of a far wider range of applications of the SSTI concept, including many in Master Data Management. The concept of whole of life visualisation applies as readily to an engineered entity as to a human life. Indeed, many engineered systems have life spans which are as long, or longer than the longest human life.

This poses many challenges in master data management as the life of the entity evolves from the initial concept, through the technical drawing and test phases, to the engineered finished product, be it a ship, aircraft, railway network or oil and gas pipeline system.

Each product may in turn contain thousands or millions of individual components, each with their own critical life histories from conception to final immolation.

Each element of any engineered product will thus be associated with a huge volume of data in many different subject fields, which include:

• Original drawings and design elements;
• Test data for the component by itself and within the engineered system;
• Ownership of the designs, of the components and of the complete system;
• Manufacturing, including the prime assembler and one or many part suppliers;
• The legislative environment in each jurisdiction in which the system and the components are used;
• Wear and tear characteristics, signs of failure and actual failure;
• “Off-label use” of components and systems outwith the original intent and specifications, and so on…

It is self evident that anyone who comes late to the party will often face huge problems in understanding all aspects of the system which they inherit for ownership, maintenance and/or replacement purposes, and the documentation and the costs associated with information management around complex systems can be enormous [4].

The questions therefore arise as to:

• Whether a data model which resembles the UHS Lifelines model would establish a valuable role across the engineered systems universe;
• Whether such a model would be a natural home for structured master data across the whole life of the component or system;
• Whether such a data model could be standardised across the relevant industries, such that data and metadata were portable and transferrable across standard interfaces and taxonomies for the whole life of the entity;
• Whether the costs of researching and implementing such a system could be estimated and matched to the projected benefits at scale;
• Whether existing organisations, and in particular the International Organisation for Standards could be co-opted into participation and partnership in such a system
• Whether and how suitable and persuasive exemplars can be developed at pace and scale;
• How and where the core data would be stored; and
• Whether the application of a mandated “whole of life” core data set and appropriate data visualisation systems could progressively be applied to all engineered components and systems, for the benefit of owners, users and maintainers throughout the life of engineered systems and components.

In summary, the Stacked, Synchronised Timeline and Iconographic (SSTI) structure, as exemplified by UHS Lifelines, would appear to offer significant potential for further development as a tool for the visualisation, interrogation and whole-of-life Master Data Management of engineered components and entities. The challenge is how to get this message out across the engineering community, and to develop practical applications.

I hope to return to these matters in future blog posts as the story evolves.

References

[1] Hales A, Cable D, Crossley E, Finlay C and Rew DA. The Design and Implementation of the Stacked, Synchronised and Iconographic Timeline Structured Electronic Patient Record in a UK NHS Global Digital Exemplar Hospital. BMJ Health & Care Informatics 2019 – https://informatics.bmj.com/content/26/1/e100025

[2] Plaisant, C., Shneiderman, B., Mushlin, R. An Information Architecture to Support the Visualization of Personal Histories.Information Processing & Management, 1997;  34: 5, pp. 581-597, 1998.

[3] Shneiderman B. The Eyes Have It: A Task by Data Type Taxonomy for Information Visualizations Proceedings  IEEE Symposium on Visual Languages  1996; 336-343

[4] Rogoway T. Boeing Is Being Paid $84 Million Just For Manuals For New Air Force One Jet – The War Zone website, April 15th 2020 – https://www.thedrive.com/the-war-zone/33034/just-manuals-from-boeing-for-new-air-force-one-jets-cost-a-whopping-84-million

The media is taking lots of data from various sources and turning it into information in an effort to help the general public understand what is happening. However, a number of media outlets are not reporting the definition of a particular figure accurately. As an example, the identical single figure for deaths is being reported in the same publications as being both “from” COVID-19, and “with” COVID-19. Two entirely different concepts.

When this pandemic is over there will be lots of people trying to make sense of the data. One of the likely measures will be deaths per 1m of the population. Comparing total deaths to a fixed measure of the population will give a better sense of the effects than simply using the figure for total deaths by itself.

This article is not attempting to analyse the current data. The purpose of the article is to explain why looking at a single figure and drawing conclusions is not giving the complete picture, why understanding the definitions behind each data element is important, and, in the case of England and Wales, how the data is collected.

Is the data fit for purpose?

Fortunately, in England and Wales the Office of National Statistics (ONS) follows good practice.

In data standards we talk about the quality of data being defined as its “fitness for purpose”. More specifically in the case of ONS data, it is the fitness for purpose with regards to the European Statistical System dimensions of quality:

• relevance – the degree to which a statistical product meets user needs in terms of content and coverage;
• accuracy and reliability – how close the estimated value in the output is to the true result;
• timeliness and punctuality: the time between the date of publication and the date to which the data refers, and the time between the actual publication and the planned publication of a statistic;
• accessibility and clarity – the ease with which users can access data, and the quality and sufficiency of metadata, illustrations and accompanying advice:
• coherence and comparability – the degree to which data derived from different sources or methods, but that refers to the same topic, is similar, and the degree to which data can be compared over time and domain, for example, geographic level;

and two other important dimensions:

• output quality trade-offs; and
• assessment of user needs and perceptions.

The ONS issues comprehensive data sets free of charge so that detailed analysis can be carried out by third parties.

How the ONS report deaths in England and Wales

The ONS produce summary information on annual deaths. The following table lists the number of deaths each year in England and Wales between 2014 and 2018.

The general rise in these figures is something to bear in mind when reviewing the average number of deaths during this period when compared to the five-year average, and is another reason why using the number of deaths per million gives better context to the figures. According to the ONS, the population of England and Wales in mid-2014 was 57,408,600, and in mid-2018 it was 59,115,809, an increase of 1,707,209. When the mid-2018 data set was published, the ONS noted that “Since mid-2000, the population of the UK has grown by almost 7.5 million and there are 2.4 million more people aged 65 to 84 years and 489,000 more aged 85 years or over.” Annual population updates are normally published by the ONS in the last week of June.

The ONS also records “excess winter deaths”. In the 2018 to 2019 winter period (December to March), there were an estimated 23,200 EWD in England and Wales. This was substantially lower than the 49,410 EWD observed in the 2017 to 2018 winter and lower than all recent years since 2013 to 2014 when there were 17,280 EWD.

How the ONS is reporting deaths involving COVID-19

As a result of the current pandemic, the ONS currently provides a separate breakdown of the numbers of deaths involving COVID-19. That is, where COVID-19 or suspected COVID-19 was mentioned anywhere on the death certificate, including in combination with other health conditions. If a death mentions COVID-19, it will not always be the main cause of death, it will sometimes be a contributary factor. The conditions mentioned on the death certificate are used to derive an underlying cause of death.

Mortality statistics in England and Wales are derived from the registration of deaths certified by a doctor or a coroner. Deriving conditions from a death certificate introduces a known variable; the accuracy of the data drawn from the certificate is dependent on the doctor completing the certificate. Before submitting a death registration through the Registration Online (RON) system, the registrar will verify that all the information provided has been entered accurately. There are some automatic validation checks within RON to help the registrar with this process. The cause of death reported represents the final underlying cause of death. This takes account of additional information received from medical practitioners or coroners after the death has been registered.

The authoritative source in England and Wales for the certification of births, marriages, and deaths is the General Register Office (GRO). The GRO pass death registration information to the ONS electronically, and the ONS: codes; compiles; and publishes these figures weekly.

The ONS short list for cause of death is based on a standard tabulation list developed in consultation with the Department of Health. This list of over 100 conditions was based on the following:

• all conditions given in the World Health Organization (WHO) basic tabulation list; with the exception of a few conditions that are so rare as certified causes of death in England and Wales that they could safely be excluded from the list;
• totals for each International Classification of Diseases, Tenth Revision (ICD-10) chapter;
• conditions used in monitoring public health targets;
• other conditions often cited by ONS.

Currently, in the UK The Department of Health and Social Care (DHSC) release daily updates on the GOV.UK website counting the total number of deaths reported to them that have occurred in hospitals among patients who have tested positive for the coronavirus (COVID-19) up until 5pm the day before.

Since 2 April, NHS England have been releasing daily updates of deaths in hospitals among patients who have tested positive for COVID-19 in England, which includes updates on previous days numbers.

The Office for National Statistics (ONS) provides figures based on all deaths registered involving COVID-19 according to death certification, whether in or out of hospital settings.

At the time of publication, further work is in progress across government to reconcile all sources of COVID-19 deaths data, but as you can see these figures are not directly comparable.

The figures produced by the ONS are about two to three weeks behind the daily reported deaths.

There has been a lot of speculation regarding the number of deaths in care homes. The ONS publishes this data.

There has also been speculation regarding whether influenza and pneumonia deaths are recorded each year. They are:

One point to note here is that the reporting of deaths involving COVID-19 is more comprehensive than the previous reporting of respiratory infections, so any comparison of these figures should make this point.

The England and Wales ONS is very clear on how it compiles the information it publishes. Authoritative data sources are always the best places to go start your search for data. Authoritative sources are normally very clear about the definitions for the data they publish, and very open about the methodologies they use. I was visiting the Korean Data Agency in Seoul at the start of this outbreak, and saw first hand how they operate. The ONS is very open, their website (www.ons.gov.uk) is a mine of useful information, and much of the text in this article describing how the numbers have been derived has been drawn from various parts of that site.

Social statisticians studying this mortality figures from pandemic will have lots of data to analyse in the future. They will also look at comparing population density rates as well as the number of deaths per million. They will also adjust the figures for the variability of testing – in both quantity and quality of the tests. The total mortality figures you read about now will have more context in the future.

How is COVID-19 data recorded where you are?

I would be very interested to hear from my fellow data experts in other countries, as to how their deaths are being recorded and published by their data agencies, and if their officially published figures for COVID-19 include deaths “involving COVID-19” where COVID-19 is not the main cause of death. Looking at the deaths per million of population figures published to date, there are some interesting variances that may be explained by the way the data is collected.

When the data is analysed in the coming years, we will need detailed quality data with clear definitions. There will be lots of variances, certain types of deaths will rise, others will fall. We will need to make sure we are comparing like numbers before we draw conclusions. This is vital to enable accurate decisions to be made about managing, or even preventing future pandemics.