Estimating the Length of Social Care Cases

A social care organisation employing 2,000 case workers believed its standard estimates of the number of hours required to complete a case were out of date. This view was also shared by the case workers themselves and their trade unions.

There was an acceptance on all sides that the accuracy of budget and resource planning were being compromised by these out-of-date estimates. A committee consisting of managers, staff and unions agreed that a detailed survey of the time case workers were spending on 9 different types of cases was needed.

Unfortunately, there was a major difficulty in that whilst a worker may only spend 100 hours on a single case, the date the case was opened and the date the case was closed could be up anything from 1 week to 3 years apart. There was no way a case lifetime survey could run for 3 years and the committee realised they needed a statistician to advise how to overcome this issue.

Why the Company Came To Me & What I Did For Them

They knew that I had the expertise to find answers and the ability to present results in a way that everyone could understand and make use of them and so they approached me.

After establishing what the objectives of the survey and analysing some existing data on case lifetimes, I recommended that the survey should aim to collect data so that 2 different methods of estimating case lifetimes.

The first survey form would ask workers to record their time in 10 minute blocks by the nature of work done and the case it related to over a two week period. The second survey form asked staff to estimate the time spent on each open case prior to the survey and the time required to complete a case if it was still open after the survey. After conducting a pilot study, I was able to estimate the likely response rate and the sample size needed.

With the data from the first form, I was able to estimate the distribution of hours spent per case type per day. This could then be multiplied by the distribution of the number of working days per case type to arrive at an estimated case lifetime and confidence interval.

For the data from the second form, the estimated time could be calibrated against a subset of workers who had been required to record their time in detail in the past. Both estimates were then combined to produce a final estimate.

What Happened Next

The committee had specified that they needed estimates for the two most important case types A & G with 95% confidence intervals within +/- 10% so they were delighted that this had been achieved. A post-survey review identified a number of ways to improve response rates and to simplify the recording form itself. I also decided that the estimation method based on the first survey form was sufficiently robust that it would not be necessary to run the second survey form as well when the survey was repeated in 2010.