Interpreting a Screening Pilot for Type 2 Diabetes in a Semi-Urban District
A district health office in a governorate of 480,000 people pilots opportunistic screening for type 2 diabetes at 6 primary care centres over 3 months. Adults aged 40-65 attending for any reason are offered a random capillary glucose; those ≥11.1 mmol/L are referred for fasting plasma glucose (FPG). Of 12,000 eligible attenders, 9,600 accept. 1,152 screen positive on capillary testing; 960 of these complete FPG, of whom 384 have FPG ≥7.0 mmol/L. A validation sub-study repeats FPG in 400 randomly selected screen-negative attenders; 8 are found to have diabetes. The programme manager proposes scaling the same algorithm to all 42 centres next year and reporting 'diabetes detected' as the sole indicator. The regional budget can fund either full scale-up or a two-stage confirmatory pathway with HbA1c for discordant cases, not both.
1. Calculate and interpret the yield and the completeness of the current pathway. What does the loss between steps tell you before any scale-up decision?
Read the worked answer
Step yields
- Uptake: 9,600/12,000 = 80%.
- Capillary positives: 1,152/9,600 = 12% of tested.
- Confirmation completion: 960/1,152 = 83.3% - 192 people with a positive initial test left the pathway without a diagnosis. This is the single largest avoidable loss and it is not random: non-completion usually clusters in shift workers, men, and people with long travel distances, so the detected cases will be biased toward those who can return.
- Confirmed among completers: 384/960 = 40% - a high false-positive burden from random capillary testing (non-fasting, post-prandial, stress hyperglycaemia).
- Programme 'detected' figure: 384 per 9,600 screened = 4%.
Undetected disease
- Screen-negative diabetes: 8/400 = 2%. Applied to the 8,448 true screen-negatives (9,600 − 1,152), that is roughly 169 missed cases - comparable to the 384 found. So the algorithm finds about two-thirds of prevalent undiagnosed diabetes and misses a third.
Implication: reporting only 'cases detected' rewards a pathway that loses 17% of positives and misses ~30% of disease. Scale-up on the current design would multiply these defects across 42 centres.
2. The manager argues that scaling up will 'find more cases, so it must be worth it'. Construct the counter-argument using the numbers, and state which design choice you would fund.
Read the worked answer
Counter-argument
- Scaling to 42 centres multiplies workload, not performance: the 17% non-completion and 40% false-positive rate scale proportionally. At 7× volume you would generate roughly 1,344 unconfirmed positives per cycle, each needing recall, counselling, and repeat visits - a queue that will lengthen and further depress completion.
- The false-positive rate has a real cost: anxiety, repeat testing, and clinician time, with no diagnosis in 60% of those recalled.
- 'More cases' is not the same as 'more benefit'. A detected case only produces health gain if it enters treatment and is retained; the current pathway does not measure treatment initiation or retention, so the indicator is uninterpretable as a measure of programme success.
Design choice Fund the two-stage confirmatory pathway, not the volume scale-up. Concretely: (1) replace random capillary glucose with a validated risk score or fasting/HbA1c-based algorithm to cut the 40% false-positive burden; (2) add same-visit or point-of-care confirmation to remove the return-visit loss; (3) use HbA1c for discordant cases to resolve the 2% residual missed disease. Then scale.
Indicators to report instead of 'cases detected': positivity rate, confirmation completion rate, PPV, treatment initiation within 3 months, and retention at 12 months - with the screen-negative validation repeated periodically to monitor sensitivity drift.