Clinical outcomes by OC adoption
Diabetes care processes · Cancer detection · ~6,100 English practices · by OC tertile and deprivation
Design: Cross-sectional comparison of ~6,100 practices with OC, IMD 2025 and outcome data, grouped into tertiles by OC rate (Feb 2026).
Diabetes data: National Diabetes Audit 2024–25 — Type 2 patients — “All 8 care processes” completed (HbA1c, BP, cholesterol, creatinine, urine albumin, foot surveillance, BMI, smoking). Note: retinal screening is excluded from 2024–25 due to data collection issues, so this is 8 not 9 processes.
Cancer data: Fingertips Cancer Services — detection rate (% of all new cancer cases treated that resulted from an urgent suspected cancer referral) — 5-year combined (2020/21–24/25).
3 treatment targets: HbA1c ≤58 mmol/mol, BP <140/80, cholesterol <5 mmol/L (Type 2 diabetes, NDA 2024–25).
IMD quintiles: Q1 = least deprived, Q5 = most deprived.
1. Diabetes: 8 care processes by OC tertile
% of Type 2 patients receiving all 8 care processes
National Diabetes Audit 2024–25
Reading: There is no meaningful difference across OC tertiles: Low 57.4%, Mid 58.7%, High 57.9%. Digital triage adoption is not associated with better or worse diabetes care process completion. Deprivation has a modest effect (least deprived ~60% vs most deprived ~54%) but this gradient is consistent across all three OC groups.
2. Diabetes: stratified by deprivation
8 care processes (%) — by OC tertile and IMD quintile
Q1 = least deprived · Q5 = most deprived
Reading: Within every deprivation quintile, the three OC groups perform similarly. The dominant pattern is the deprivation gradient: practices serving more deprived populations complete fewer care processes regardless of OC intensity. OC adoption neither helps nor hinders chronic disease management as measured here.
3. Diabetes: 3 treatment targets by OC tertile
% of Type 2 patients achieving all 3 treatment targets
HbA1c ≤58, BP <140/80, cholesterol <5 · NDA 2024–25
Reading: There is a small but consistent gradient: Low OC 46.2%, Mid 45.7%, High 45.1%. Higher OC adoption is associated with slightly lower treatment target achievement, though the absolute difference is just over 1 percentage point. After adjusting for deprivation in linear regression, the association remains statistically significant (β = −0.004 per unit OC rate, p < 0.001) but explains very little variance (R² = 0.014).
4. 3 treatment targets: stratified by deprivation
3 treatment targets (%) — by OC tertile and IMD quintile
Q1 = least deprived · Q5 = most deprived
Reading: The slight OC gradient is consistent within every deprivation quintile: Low OC practices achieve ~0.5–1.5pp more than High OC practices. The deprivation gradient is also modest (Q1 ~47% vs Q5 ~45%). The pattern is real but small, and cross-sectional data cannot tell us whether OC drives lower target achievement or whether practices with other characteristics (e.g. younger populations, more acute demand) both adopt OC more and achieve fewer targets.
5. Cancer detection rate by OC tertile
% of new cancers detected via urgent suspected cancer referral
Fingertips Cancer Services · 5-year combined 2020/21–24/25
Reading: Cancer detection rates are similar across OC tertiles overall: Low 53.0%, Mid 53.8%, High 54.1%. The difference between Low and High is ~1 percentage point — statistically significant given the large sample sizes, but of uncertain clinical importance.
6. Cancer detection: stratified by deprivation
Detection rate (%) — by OC tertile and IMD quintile
Q1 = least deprived · Q5 = most deprived
Reading: Deprivation is the dominant driver: detection rates fall from ~57% in the least deprived practices to ~49% in the most deprived, a gradient of 8 percentage points. Within Q1–Q4, the OC groups are near-identical. However, in Q5 (most deprived), High OC practices detect 50.4% of cancers via urgent referral compared with 48.5% for Low OC (p < 0.001). This is the only quintile where a statistically significant difference emerges. It may suggest that digital triage helps maintain referral access in the most deprived areas, though cross-sectional data cannot establish causation.
7. Linear regression
Adjusted associations: OC rate, IMD score and list size as predictors
OC rate = submissions per 1,000 patients · IMD score = continuous deprivation · List size in thousands
| Outcome | OC rate β | p | IMD β | p | List size β | p | R² |
|---|
| 8 care processes | −0.001 | 0.569 | −0.119 | <0.001 | 0.055 | 0.060 | 0.009 |
| 3 treatment targets | −0.003 | <0.001 | −0.062 | <0.001 | −0.099 | <0.001 | 0.025 |
| Cancer detection | 0.001 | 0.230 | −0.232 | <0.001 | 0.030 | 0.009 | 0.165 |
| Cancer (+ OC×IMD) | −0.004 | 0.018 | −0.254 | <0.001 | 0.031 | 0.008 | 0.166 |
Reading: After adjusting for deprivation and practice list size, OC rate is not a significant predictor of 8 care process completion (p = 0.57). For 3 treatment targets, OC rate retains a small negative association (β = −0.003, p < 0.001), and larger practices also achieve fewer targets (β = −0.099 per 1,000 patients, p < 0.001). For cancer detection, the OC main effect is not significant (p = 0.23) but the OC × IMD interaction is (p = 0.001): in more deprived areas, higher OC adoption is associated with modestly higher cancer detection. Across all models, deprivation is the dominant predictor, and R² values are low — most variation in outcomes is explained by factors not captured here.
8. Does OC narrow the deprivation gap?
Q1–Q5 gap (pp) by OC tertile — the inverse care law
Gap = mean in Q1 (least deprived) minus mean in Q5 (most deprived) · smaller bar = less inequality · error bars = 95% CI (bootstrap)
Reading: For cancer detection, the deprivation gap narrows significantly with OC adoption: 8.3pp in Low OC vs 6.1pp in High OC (a reduction of 2.2pp). The OC × IMD interaction confirms this formally (p = 0.001). For diabetes care processes and treatment targets, the gap is essentially unchanged across OC tertiles. This suggests that digital triage may help reduce health inequality in access-sensitive outcomes (getting referred for cancer) without measurably affecting inequality in chronic disease management.
9. Summary
Outcome measures by OC tertile
| 8 care processes (%) | 3 treatment targets (%) | Cancer detection (%) |
|---|
| Low OC | 57.4 | 46.2 | 53.0 |
| Mid OC | 58.7 | 45.7 | 53.8 |
| High OC | 57.9 | 45.1 | 54.1 |
Key finding: On the measurable clinical outcomes available at practice level, OC adoption makes little discernible difference. Diabetes care processes and treatment targets are essentially flat across OC tertiles, with deprivation the dominant predictor. Cancer detection shows one notable exception: in Q5 (most deprived), High OC practices have a statistically significantly higher detection rate than Low OC practices (50.4% vs 48.5%, p < 0.001). The 3 treatment targets show a very slight reverse gradient (Low OC 46.2% vs High OC 45.1%). Cross-sectional data cannot establish causation in either case.
What can we measure? Patient satisfaction shows a clear gradient with OC adoption — lower at high OC. Demand shows a clear gradient — higher at high OC. Clinical outcomes show mostly no association, with one exception: in the most deprived practices, higher OC adoption is associated with modestly higher cancer detection rates via urgent referral. This could reflect better access to referral pathways, but cross-sectional data cannot tell us whether OC is the cause. The measures we can access — care process completion, treatment targets, cancer detection — are relatively blunt and may not be sensitive to the kind of changes digital triage could plausibly produce. Finer-grained outcome data (ambulatory care sensitive emergency admissions, continuity of care) is not currently published at practice level. Until it is, the outcome question remains largely open.