## Clinical Decision Support (2006)

### BibTeX

@MISC{Cumming06clinicaldecision,

author = {Jonathan Cumming and Jonathan Cumming},

title = {Clinical Decision Support},

year = {2006}

}

### OpenURL

### Abstract

Within orthopædics, clinicians routinely take multiple measurements on patients during the course of their treatment, often repeating the same measurements before and after operations, and subsequently at periodic follow-up consultations. This data combined with additional factors gives a wealth of information, resulting in a high-dimensional data set with a mixture of data types and a longitudinal aspect; all of which can be problematic in statistical analysis. Therefore, general statistical methods for the investigation and analysis of a generic medical data set are presented and developed. Methods are proposed for supporting exploratory analysis of the data via novel visualisations of the patient’s status over time across multiple variables, thus giving an easily interpretable overview of this evolution. To address the problem of high dimensionality of the data, a new approach to variable selection is proposed and developed using principal variables. The method is further extended by the use of temporal smoothing to tackle data with this repeated measures aspect allowing for the simultaneous reduction of the patient status variables over time. The ultimate goal of these analyses is to determine an appropriate model for the orthopædic data, with a focus on the modelling of the time series of patient progress. The techniques of graphical modelling and, in particular, those of chain graphs lend themselves to this problem. Additionally, they have the added benefit of a simple and intuitive visualisation which is of benefit to clinicians. All of these methods are illustrated via their application to two large-scale case study data sets concerning total joint replacement.

### Citations

1357 |
The Visual Display of Quantitative Information
- Tufte
- 1983
(Show Context)
Citation Context ...e. The goal of EDA is to discover patterns in, and develop an understanding of, the data whilst generating hypotheses for subsequent investigation. The focus is typically on the use of visualisations =-=[19, 119]-=- and robust methods rather than complex model building. Tukey described EDA as a ‘foundation stone’ [121] in the analytic process. Therefore, this chapter contains an exploratory analysis of the two k... |

1155 |
Exploratory Data Analysis
- Tukey
- 1977
(Show Context)
Citation Context ...3: Summary of several variables of the hips data set.sChapter 3 Exploratory Data Analysis Exploratory Data Analysis (EDA) is a well-established statistical approach which stems from the work of Tukey =-=[121]-=- and is well documented in the literature. The goal of EDA is to discover patterns in, and develop an understanding of, the data whilst generating hypotheses for subsequent investigation. The focus is... |

782 |
Envisioning Information
- Tufte
- 1990
(Show Context)
Citation Context ...eans for NHS and private patients.sChapter 4 Visualisations The use of graphical methods to present data and to communicate the statistical features thereof has a long history with Tukey [121], Tufte =-=[119, 120]-=-, Cleveland [18, 19], and more recently Wilkinson [126] having written extensively on the subject. The usefulness of such methods is undisputed; as Tufte says: ‘Graphics reveal data’. By showing a gre... |

435 |
Estimating the dimensions of a model
- Schwartz
- 1978
(Show Context)
Citation Context ... all variance matrices were considered to be equal was considered to be unreasonable and tests of this assumption corroborated this. The edge selection process used the Bayesian Information Criterion =-=[112]-=- rather than the standard χ 2 tests. The motivation for this is that at any stage in the development of the model, a number of highly significant edges will be eligible for inclusion in the model. Due... |

170 |
Correlation and causation
- Wright
- 1921
(Show Context)
Citation Context ... in which each variable is a node and each dependency is an edge. Graphical modelling as a statistical methodology can be traced back to the early 20th century with foundations in statistical physics =-=[128]-=- and path analysis [54]. However, the major developments in the field are of a more recent origin with notable contributions by Lauritzen, Wermuth and Cox [25, 124, 22, 83]. The goal of this chapter i... |

108 | The Grammar of Graphics - Wilkinson - 1999 |

101 |
Using multivariate statistics
- Tabachnik, Videll
- 1989
(Show Context)
Citation Context ...f differences and perform multiple comparisons. They applied their results to illustrate the main effects of several ANOVA models. Another near-identical approach can be found within profile analysis =-=[57, 117, 114]-=- which is a specialisation of multivariate analysis of variance (MANOVA) to a situation with multiple responses all on a similar scale. A further extension where there are several responses observed a... |

66 |
Cross validatory estimation of the number of components in factor and principal component analysis, Technometrics 20
- Wold
- 1978
(Show Context)
Citation Context ...mining the effective dimensionality where one seeks the PC beyond which the variances decrease in a linear fashion - the PCs beyond this point are regarded as representing noise within the data. Wold =-=[127]-=- and Eastment and Krzanowski [36] propose a cross-validatory approach for determining the number of components to retain. Eastment and Krzanowski’s approach is based on successively predicting each el... |

57 |
On substantive research hypotheses, conditional independence graphs and graphical chain models
- vVermuth, Lauritzen
- 1990
(Show Context)
Citation Context ...with foundations in statistical physics [128] and path analysis [54]. However, the major developments in the field are of a more recent origin with notable contributions by Lauritzen, Wermuth and Cox =-=[25, 124, 22, 83]-=-. The goal of this chapter is twofold: first to provide a brief overview of the fundamentals of the graphical modelling methodology; and secondly to apply these methods in an exploratory fashion to ou... |

35 |
Determining the number of components from the matrix of partial correlations
- Velicer
- 1976
(Show Context)
Citation Context ...idation. Both methods then consider the prediction error sum of squares (PRESS) for different numbers of components and construct statistics based on functions of the various values of PRESS. Velicer =-=[122]-=- adopted an approach based on partial correlation in order to determine the number of components. He proposed considering the partial correlations between p variables given the first m principal compo... |

21 | On chain graph models for description of conditional independence structures
- Studeny, Bouckaert
- 1998
(Show Context)
Citation Context ...om x to y. If x → y, y → x or x ↔ y then we say that x and y are adjacent (x ∼ y). These graphs are known as block-recursive or chain graphs and their properties are well documented in the literature =-=[83, 124, 52, 22, 116, 9, 82]-=-. The class of chain graphs includes both undirected graphs and directed acyclic graphs (DAGs) as special cases when all edges are undirected or directed respectively. One of the restrictions of this ... |

20 |
trials of medical decision-aids: potential problems and solutions
- Wyatt, Field
(Show Context)
Citation Context ...itoring of ventilator systems in intensive care, and further to using toxicological information to warn clinicians if they have prescribed a dangerous drug combination. It was Wyatt and Spiegelhalter =-=[129]-=- who gave the now commonly accepted definition of a clinical decision support system as: “an active knowledge system which uses two or more items of patient data to generate case-specific advice.” Fro... |

16 |
Methods for analysis of longitudinal data: blood lead concentrations and cognitive development
- Waternaux, Laird, et al.
- 1989
(Show Context)
Citation Context ...e the set of key variables measuring the patient’s status. In fact, these replicated observations of the same variables fall into the framework for repeated measures [24, 31, 86] or longitudinal data =-=[32, 123]-=-. However these frameworks would struggle with the mixture of data types and the inclusion of large amounts of extraneous information that is not repeatedly observed.sPatient Demographics Initial Cons... |

8 |
Graphical Models
- Whittaker
- 1990
(Show Context)
Citation Context ...aphical models themselves and model selection methods. This section is intended only as an overview - a more comprehensive treatise on the subject can be found in the books by Edwards [38], Whittaker =-=[125]-=-, or finally Lauritzen [81] who gives a thorough exposition of the theoretical foundations of the methodology. Some applications of these methods available in the literature are also presented. The se... |

3 |
Data analytic techniques for treatment outcome studies with pretest/posttest measurements: an extensive primer
- Sheeber, Sorensen, et al.
- 1996
(Show Context)
Citation Context ...f differences and perform multiple comparisons. They applied their results to illustrate the main effects of several ANOVA models. Another near-identical approach can be found within profile analysis =-=[57, 117, 114]-=- which is a specialisation of multivariate analysis of variance (MANOVA) to a situation with multiple responses all on a similar scale. A further extension where there are several responses observed a... |

1 |
Encyclopedia Of Artifical Intelligence
- Shapiro
- 1990
(Show Context)
Citation Context ...on to the variable selection process. Thus the simple stepwise search could be improved upon by considering a tree-based search strategy. The application of simple search algorithms such as A* search =-=[113]-=- would likely improve the quality of the returned subset, at the expense of additional computation. One advantage of using an algorithm and search strategy such as A* search is that it is guaranteed t... |

1 |
et al. Kendall’s Advanced Theory of Statistics: Classical Inference and the Linear Model, volume 2A of Kendall’s Library of Statistics. Arnold Publishers, 6th edition
- Stuart, Ord, et al.
- 1998
(Show Context)
Citation Context ...at the expression for the recursive weights in (6.15), we notice that this expression is, in fact, closely related to the multiple correlation coefficient. We have from Equation 28.56 in Stuart et al =-=[115]-=- that: 1 − R 2 1(2...p) = (1 − ρ 2 12)(1 − ρ 2 13.2)...(1 − ρ 2 1p.2...(p−1)), (6.16) where R1(2...p) is the multiple correlation coefficient between variable 1 and variabless6.4. Stepwise Selection P... |

1 |
Guide to recording information about knee replacements: a manual for use in outpatient clinics and hospitals
- Tew, Waugh
- 1980
(Show Context)
Citation Context ...and ten years post-operatively. A total of 23 measurements were recorded at each consultation to monitor patient progress. These measurements form the components of the Nottingham knee scoring system =-=[118]-=- and thus compose a small repeated measures data set, corresponding to the Measurements in the general data structure of Figure 2.3. The 23 measurements are composed of a mixture of both ordinal and c... |