Introduction

This page

Imagine your doctor tells you that your blood pressure is elevated and he prescribes medicine to help you to control it. He also suggests that you buy a blood pressure monitoring machine and take and record your blood pressure yourself on a regular basis. What follows is the case history of one patient over the course of almost a year of treatment following the initial diagnosis of elevated blood pressure.

The doctor initially prescribed 5 mg per day of Enaril that he increased to 10 mg a day on 2nd September. The patient is male, Causcasian, now 66 years old and in otherwise good health.

The Data

The data are contained in the Excel CSV file (for Word Press security reasons, it is not possible to upload the Power BI file, sorry) that you can download from the link given at the bottom of the page. I have kept the database simple:

The doctor advised taking blood pressure readings at the same or similar time every day: that was not always possible but since there is a year’s worth of data that actually provides more insights, I think.

Please also note that there are gaps in the data when the patient was away from his monitoring machine and could not record his pressure.

Statistics

In line with the way statistics can work, I have tended to work with median rather than mean values, as you can see in many of the visualisations I have prepared.

Visualisations

Here are my visualisations, including some commentary:

Firstly, the medians of the three measurements, systolic, diastolic and pulse both by time of the day and by month. Note, there is a separate time column on which I havebased the visualisation by time but for the dates, PBI shows them initially asDate: Year, Quarter, Month, Day and here I chose Month.

Is there anything odd or unusual in what you see here?

Now some Scattergraphs, below; and in the left hand column, showing one variable against another: the possible interaction or relationship between variables. There are outliers here, I am sure you can see them! Would you investigate them?

In the right hand column, I have created stacked column charts. The Diastolic by Systolic graph tells us, given the Systolic reading of, say, 120, the Count of number of Diastolic readings is 1; there are seven Diastolic readings where the systolic reading is 138. Is this useful information?

The second graph on the right shows the median pulse rate given Diastolic readings. For example, where Diastolic is 70, the median pulse rate is 65; and where the Diastolic reading is 79, the median pulse rate is 72. See the graph below that showsCount of Pulse by Diastolic next to Median of Pulse by Diastolic.

You might find the graphic that follows, showing median and counts of pulse rates by diastolic more revealing than just the median pulse rate v Diastolic:

Two tables and five graphs, now.

The table shows the raw data but I have allowed PBI to show the dates by year, by quarter, by month and by day. There is no particular value in doing that, so I have done it just to show what it might look like.

In the middle column, I have shown the variables as averages (means) by month. You can see that there appears to be no great variation by month.

In the right hand column, I have introduced a new variable: Mean Arterial Pressure (MAP). I found this variable following a search on the Web for additional metrics linked to blood pressure analysis. The formula for MAP is 1/3 * Systolic + 2/3 * Diastolic and I included it in my analysis by creating a new column for it, using this DAX formula: MAP= 1/3*dw_bp_2018[Systolic]+2/3*dw_bp_2018[Diastolic]

My analysis of MAP

Median MAP by month, top graph

Median MAP by time and for this line graph, I have added two constant lines: at 70 and at 110. These two values are the lower and upper boundaries set by physicians as the limits that a healthy person should not be outside of.

The table shows the MAP by date and time.

The final page shows some Cards, a table, two line graphs and two scattergraphs:

The cards are simply showing median values of Systolic, Diastolic and MAP I did not add Pulse but you could easily do that.

The table shows the Median MAP by month and since the data here cover almost an entire year, you can see how this patient’s arterial pressure has been progressing from January to December.

The top two graphs on the right hand side are scattergraphs and they show the median MAP by day number (ie day of the month) and time: are they of interest? By day, medianMAP is scattered and by time, it seems to increase as the day goes on, starting at around 100 in the morning and tending towards 106 or 107 by bed time.

The two line graphs show the relationships between median MAP and Systolic (X axis) then Diastolic (X axis).

Conclusions

Using just five variables, date, time, systolic, diastolic and pulse rates and deriving a sixth, we can see that PBI can provide us with a relatively sophisticated view of the blood pressure history of this patient.

Of course, there are more analyses that we could have done but did not: correlations, descriptive statistics and so on and you are free to work with them.

I have made no mention in my analysis of the change of dosage of the medicine and the age/sex profile of the patient. We should.

UPDATE

Since I published this page, I have been working on Measures in my PBIX file, more particularly, New Quick Measures. So far, I have drawn no conclusions as to how useful they are and how meaningful they are. Here is an example, however, of one page from my updated report:

Excel file of data only … uploading PBI files here is not possible. 

Duncan Williamson

15th December 2018

%d bloggers like this: