__Stata ados__

__Minimally important change (MIC) calculator__

ROCMIC estimates minimally important change (MIC) thresholds using three different methods. The first is the cut-point corresponding to a 45 degree tangent line intersection; this is mathematically equivalent to the point at which the sensitivity and specificity are closest together. The second is the cut-point corresponding to the smallest sum of 1-sensitivity and 1-specificity; this methodology has been proposed by researchers from the EMGO Institute. The third is the cut-point corresponding to the smallest sum of squares of 1-sensitivity and 1-specificity in accordance with Pythagoras' theorem, as has been recommended by Froud and Abel. These programs work in conjunction with Stata's **bootstrap** command, allowing calculation of a confidence interval around the MIC point-estimate.

See:

**Froud R, Abel, G. Using ROC curves to choose minimally important change thresholds when sensitivity and specificity are valued equally: the forgotten lesson of Pythagoras. Theoretical considerations and an example application of change in health status. PLOS One 2014; DOI: 10.1371/journal.pone.0114468**

**Farrar JT, Young JP, Jr., LaMoreaux L, Werth JL, Poole RM. Clinical importance of changes in chronic pain intensity measured on an 11-point numerical pain rating scale. Pain 2001;94(2):149-58.**

**de Vet H, Terluin B, Knol D, Roorda L, Mokkink B, Ostelo R, et al. There are three different ways to quantify the uncertainty when 'minimally important change' (MIC) values are applied to individual patients. J Clin Epidemiol 2009**

**Installation instructions:**

*rocmic is designed for Stata version 10.1 and higher. If you have an older version of Stata, please contact me.* *

Open Stata and type *ssc install rocmic*

Type ** help rocmic** for help

Alternatively click here for installation files

If you use this software module please reference it as follows in any relevant scientific reports:

Froud, R. Abel, G. 2014 ROCMIC v2.0.0 Stata module to estimate minimally important change (MIC) thresholds for continuous clinical outcome measures using ROC curves, Statistical Software Components S457052, Boston College Department of Economics. RePEc:boc:bocode:s457052

**If you are familiar with Stata programs you may alter the version number in the ado file and try the program. However, unless the current bootstrap ado( v4.3.9) is installed, the program runs incorrectly in conjunction with the bootstrap program. *

__Number needed to treat (NNT) calculator, with confidence interval derived from Wilson scores__

These programs estimate(named 'calculator' as people don't tend to search for estimators) NNT and calculate Newcombe Method 10 Confidence Intervals (CIs) for the risk difference of improvement, or benefit (Improvements gained and deteriorations prevented), between clinical trial groups, then reciprocally transforms the results giving NNTs with upper and lower CIs. Where CIs are calculated at all for NNTs, they tend to be Wald intervals. These have poor coverage properties and are prone to aberrations. Confidence intervals derived from Wilson scores are computationally more complicated, but have better coverage properties. These CIs are calculated by **bcii**. In **bcib**, I have modified Newcombe's Method 10 to incorporate the extra variance term introduced by considering deteriorations prevented as well as improvements gained. Further details can be found in the paper below.

Open Stata and type: *ssc install bcii*

Type:** help bcii **or

**for help**

*help bcib*Alternatively click here for installation files

If you use this module please include the following in scientific reports:

**Froud, R, Eldridge, S, Lall, R, Underwood, M. Estimating Number Needed to Treat from continuous outcomes in randomised controlled trials: methodological challenges and a worked example using data from the UK Back pain Exercise and Manipulation (BEAM) trial. BMC Medical Research Methodology 2009, 9:35**

Froud, R. BCII: Stata module to to estimate the number needed to treat (NNT) and confidence intervals for patients improving, or 'benefiting' in a randomised controlled trial, Statistical Software Components S457053, Boston College Department of Economics, 2009

__Executable programs__

Clinistat v0.2.exe (901Kb)

(Contains Samclus)

Written with Sandra Eldridge

This program will calculate the sample size required for a cluster randomised trial if you want to know how many clusters to recruit and have the following:

1) An estimate of the mean cluster size

2) An estimate of the intra-cluster correlation coefficient

3) Either:

• An estimate of the coefficient of variation of cluster size (standard deviation of cluster sizes divided by the mean cluster size), or

• An idea of the smallest and largest cluster size that you expect in the trial (if cluster sizes will be equal then enter zero as the coefficient of variation)

Results will appear in the dialogue box, but also can be saved in a text file on your desktop if required.

Note that when cluster sizes vary the sample size required depends on the type of analysis to be undertaken, and the estimate of sample size provided by this program is conservative. Methods are described in: **Eldridge SM, Ashby D, Kerry S. Sample size for cluster randomized trials: effect of coefficient of variation of cluster size and analysis method. Int J Epidemiol 2006; 35(5):1292-1300.**

We are trialing SAMCLUS at the moment, so if you experience any problems using this program or have suggestions for its improvement, please let me know ...

- nb this version of Clinistat contains only SAMCLUS.** It will not run correctly on Windows Vista (Vista users please use EXCEL version below).** I will eventually get round to finishing v1, which includes a suite of programs & is Vista compatible. I note that this version of SAMCLUS does not round up estimates & it is not possible to have a fraction of a person! (n.b. *The Android version is more up-to-date*)

__Microsoft EXCEL programs__

SDC LOA Calc v0.1.xls (538Kb)

Written with Caroline Terwee

1) Smallest Detectable Change (SDC) calculator (a.k.a Minimally Detectable Change (MDC))

This program will calculate the standard error of measurement (SEM), Intraclass Correlation Coefficient (ICC), and SDC on an individual level and group level, for a health related quality of life instrument. It requires the following information:

Mean Sums of sQuares (MSQ) for:

1) Patients (between patient variance)

2) Time or observers (within patient variance)

3) The residual error

These MSQs can be calculated easily in SPSS using a repeated measures ANOVA. Results will appear in the dialogue box, but will also be stored within the spreadsheet itself.

Methods are described in:

Streiner D and Norman GR, Health measurement scales. A practical guide to their development and use. 2006, New York: Oxford university press.

**de Vet HCW, Terwee CB, Knol DL, Bouter LM. When to use agreement versus reliability measures. J Clin Epidemiol 2006;59:1033-1039.**

2) Limits of Agreement (LOA) calculator

This program calculates limits of agreement (LOA) as described by Bland and Altman.

Methods are described in:

**Bland, J and Altman, DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet, 1986; i: 307-10.**

This program was adapted by Robert Froud from a spreadsheet written by Caroline Terwee. For comments and / or suggestions please contact either: cb.terwee@vumc.nl or r.j.froud@qmul.ac.uk. *This will be incorporated in Clinistat v1.0*

If the program does not run: Click Tools - Macro - Security - Select 'Medium' Exit EXCEL and then restart. In EXCEL 2007, macro security is found in the 'Developer' menu |

Microsoft Office 2007 Home and Student Edition (3 User Licence) (PC)

samclus v0.1.xls (305Kb)

Written with Sandra Eldridge

An older EXCEL version of SAMCLUS (described above).

If the program does not run: Click Tools - Macro - Security - Select 'Medium' Exit EXCEL and then restart. In EXCEL 2007, macro security is found in the 'Developer' menu |

OR RR calculator with CIs.xls (263Kb)