How should I approach balancing the results of this data?

[halesowenmum]

Hi

I've got a table showing %s of totals within the following categories:

• Patients who after leaving hospital had an attempted telephone contact with a nurse but who were not contactable
• Patients who after leaving hospital had contact with a nurse over the phone
• Patients who after leaving hospital had contact with a nurse over the phone and via a home visit.

I'm viewing the uncontactable lot as my Control Group against which I compare how successful the telephone and the telephone and home visit interventions were. The assumption is that if you not only call someone but you also visit them at home they should be much less likely to come back in. However, firstly the patients picked to receive a phone call were picked because it was felt they had a low to moderate risk of readmitting, and for those who had home visits the entire reason for visiting them was because the clinician knew that they were much more likely to readmit (that's why they went to their house to visit them).

So comparing the phoned only against the couldn't contact and comparing phoned and seen at home against the couldn't contact, straight, isn't a fair comparison as the home visited group is more likely to readmit right from the outset.

I took the % of figures and calculated the number greater or less than the control group for both telephoned only and phoned and home visited. This is shown on the table below:

	BK	BM	BO
	Did not readmit	Readmitted once	Readmitted more than once
	POSITIVE: Obviously not readmitting is the ideal	NEGATIVE: If you're going to readmit readmitting once only isn't too bad	NEGATIVE: If you're going to readmit, readmitting more than once is the worst
141	59.1%	31.8%	9.1%	This row contains the Control Group %s of results
143	78.8%	19.7%	1.5%	This row contains the %s of results for those telephoned only
144	1.33	0.62	0.17	Greater/lesser than control group (row 141)
146	58.8%	35.3%	5.9%	This row contains the %s of results for those phoned and visited at home
147	1.00	1.11	0.65	Greater/lesser than control group (row 141)

<colgroup><col><col><col><col><col><col span="2"></colgroup><tbody>
</tbody>

Some givens:

• At the point of initially selecting patients to call they all have roughly the same risk of readmission so before the telephone is picked up the playing field is level
• Presumably the discussion over the phone for the ones the clinician was able to speak to will slightly reduce their risk of readmitting so you might expect to see a reduction in readmissions for these
• The assumption is that those who have a phone call and a home visit must be way less likely to readmit - however as described below there are reasons why this is not the case.

Looking at the readmission rates:

• Did Not Readmit rate and Readmitted Once rate for home visited patients are virtually the same as the Control Group
• Those home visited were home visited because during the phone call it was noted that their responses and past medical history showed they were more than moderately likely to readmit (and that is why a home visit was conducted).

So my understanding therefore is that you can't compare these data 'cold' because they aren't like for like in terms of readmission risk.

I beg of whichever kindly soul takes on this question, please make your answer at the 'for Dummies' level as highly statistical analysis/data modelling equations and the like I just won't understand !

What I'd like advice on is which of the above numbers to adjust, by how much and what formula (just reduce by a particular % or by the difference or the difference x 2....).

Many thanks indeed.