The cure rate that moved nine points

The draft annual report gave a cure rate of 73.6%, below the Sphere minimum of 75%. The final report gives 82.3%, comfortably above it. No data was corrected between the two.

A donor’s monitoring adviser has noticed and asked. You have the register and an hour.

The file

cmam-admissions-2024.v1.csv — 1,100 admissions across six sites, with admission and discharge measurements, dates and outcome. Synthetic.

The task

Part one. Reconstruct all three cure rates from the register. Show which denominator produces each, and state in one line what each denominator assumes about the children it excludes or includes.

Part two. Say which one the Sphere minimum standards specify, and cite the two exclusions that define it. Then say whether the programme meets the standard, and whether the draft was wrong or merely differently defined.

Part three. Compute the full performance table on the Sphere denominator — cured, defaulted, died, non-response — by site, and flag every cell outside a Sphere threshold. The district figure passes every standard; at least one site does not.

Part four. Write the reply to the donor. Four sentences: what changed, why the final figure is the defensible one, what the programme’s performance actually is, and the one finding the district figure was hiding.

What you will need to handle

Three things, and none is a data error.

• Seventy-one children have no outcome. They were admitted late enough that they had not been discharged at the cut-off. Decide what to do and justify it.
• Forty-five children were transferred. Sphere treats these differently from the other outcomes and the reason is not arbitrary.
• Nine discharge weights are below the admission weight. Say whether they affect any figure in your table, and what you would do about them.

Three reference points

The three cure rates are approximately 73.6%, 78.7% and 82.3%. If none of yours lands on one of those, work out which denominator you built before going further.

The programme-wide defaulter rate is about 13.7%, and the worst site is above the Sphere maximum.

The questions to answer in prose

Three sentences each.

1. The draft’s denominator counted children still in treatment. Explain what that implicitly claims about those children, and why it biases the cure rate in a predictable direction.

2. Sphere excludes transfers from the performance denominator. Give the reason, and say what would happen to a district total if two programmes both counted their transfers as outcomes.

3. Your by-site table has at least one cell outside a Sphere threshold while the district figure passes. Name the site and the indicator, say what the median length of stay for defaulters suggests about the cause, and give a corrective action that is not “counselling”.

What to hand in

An R script producing a three-row denominator table, the by-site performance table with Sphere flags, and the four-sentence reply as a comment at the foot of the file. Both tables printed and written to outputs/.

How to know you are done

Your denominator table shows the excluded counts, not just the totals, so a reader can reconstruct any of the three. Every rate in your by-site table carries its n, and the flags are set against the Sphere thresholds rather than against the district mean.