Patient satisfaction by OC level

Reading: Overall satisfaction is 80.2% in Low OC practices but only 73.7% in High OC — a 6.5 percentage point gap. The proportion reporting poor experience is 9.1% in Low OC and 12.9% in High. This does not mean OC causes lower satisfaction — the practices that adopted OC most heavily may have had worse access to start with, which is why they adopted it.

Reading: In-person contact is 76% in Low OC but only 53% in High. Online contact is 3% in Low and 22% in High. Phone contact is stable at 13–14%. The difference is between in-person and online, not between phone and online. This aligns with the CBT data: phone calls are lower in high OC practices, partly replaced by OC, and a substantial proportion of contacts in low OC practices are walk-in or in-person.

Reading: Phone ease is 38% in Low OC but only 25% in High; phone difficulty is 26% in Low and 36% in High. Patients in high-OC practices find the phone harder to use, not easier. This is consistent with the CBT data showing that answer rates outside rush hour are worse in high-OC practices.

Reading: Website ease is 87% in Low OC but only 59% in High — a striking finding. Patients in high-OC practices are much more likely to have used the website, and those who use it rate it less favourably than those in low-OC practices where fewer have tried. Reception helpfulness is also lower in high OC practices (82% vs 87%).

Reading: Same-day next steps are 74% in Low OC but only 54% in High. Next-working-day is 16% in Low and 32% in High. Patients in high-OC practices wait longer for a response. This is consistent with the OC workflow: a form is submitted, processed, and a response sent — inherently slower than an immediate phone conversation, even if the form was submitted at 8am.

Reading: The experience of making contact is 76% good in Low OC but only 67% in High. Poor contact experience is 12% in Low and 19% in High. This is the sharpest gradient in the survey — the process of getting through to the practice is notably worse in high-OC practices, even though (or perhaps because) the channel has changed.

Deprivation: Low-OC practices are slightly more deprived (mean IMD 25.5) than high-OC (22.0). Deprivation does not explain the satisfaction gap — it runs the other way. List size is a stronger confounder: high-OC practices are much larger (12,565 vs 7,960 patients). Bigger practices independently predict lower satisfaction.

Age: The age profile is virtually identical across tertiles (~18.5–18.9% over 65). But older populations rate their practice more highly (r = 0.28), so age is a confounder for satisfaction generally, though not between these groups. Region: Low-OC is concentrated in the North East & Yorkshire (21%) and Midlands (22%); High-OC skews towards East of England (15%) and South East (15%). Regional differences in both OC adoption and satisfaction culture could contribute.

Reading: The OC gradient persists within every deprivation band. But the striking finding is the crossover: the least-deprived practices in the high-OC group report the same level of satisfaction (77%) as the most-deprived practices in the low-OC group (76%). Affluent practices with high OC adoption are no more satisfied than deprived practices with low OC.

Reading: OC rate, deprivation, list size, and age profile together explain 17% of the variation in satisfaction — most of what drives satisfaction lies elsewhere (staff, premises, culture, continuity). But we can decompose the 6.5 percentage-point gap between the Low and High OC groups:

Key point: OC rate accounts for 4.9 of the 6.5 percentage points. List size adds another 1.2pp. Deprivation works in the opposite direction — the high-OC group is less deprived, which should predict higher satisfaction, not lower. The practices with the highest OC use are less deprived and have similar age profiles — factors that should, if anything, give them higher satisfaction. Yet they score lower. The gap isn’t explained by who these practices serve — if anything, their populations should give them an advantage.

R² in context: These four variables explain only 17% of the variation between individual practices. The other 83% reflects things we cannot measure here — staffing, premises, culture, continuity, leadership. Any single practice’s satisfaction is shaped mostly by local factors. But when we average across 2,000 practices per tertile, that individual noise cancels out and the systematic differences come through clearly. It is at the population level that policy decisions are made, and at that level the OC signal is strong.

Method note: The regression is fitted on all 6,012 individual practices to get the most precise estimate of each coefficient (all t > 12, all p < 10⁻³⁷). We then multiply each coefficient by the difference in group means to decompose the tertile-level gap. For example: the coefficient for OC rate is −0.020 per submission per 1,000 patients; the high-OC group averages 242 more submissions than the low-OC group; so OC rate accounts for 242 × 0.020 = 4.9 percentage points of the gap. This is a standard approach (Blinder–Oaxaca decomposition).

Reading: There is no smooth gradient from Q1 to Q4 (all cluster at 72–74%). But Q5 — practices with the longest, most intensive OC history (mean rate 11.7 in Aug 2024) — stand out at 77.6%, roughly 4pp above the rest of the high-OC group. 733 practices went from zero OC in Aug 2024 to the top tertile by Feb 2026; their satisfaction (73.5%) is no worse than practices that ramped up more gradually.

Interpretation: The Q5 result is consistent with a transition effect (disruption eases with longer exposure), but it could equally reflect organisational competence — practices that adopted early and went all-in may simply be better-run. Even Q5 remains 2.6pp below the Low OC reference (80.2%), and much of that residual gap is likely explained by list size (Q5 averages 12,667 vs 7,949). ANOVA across all groups: F = 9.1, p < 0.001.

Key finding: The temporal OC measures (Aug 2024 baseline and change to Mar 2025) explain far less of the satisfaction gap than the Feb 2026 cross-sectional rate. When we use the temporal measures, OC accounts for only 1.1pp of the 6.5pp gap, and 4.5pp is left unexplained. This suggests that the Feb 2026 OC rate is acting as a proxy for a bundle of practice characteristics — not just OC use, but everything that correlates with being a high-OC practice.

But the OC signal is real. In Model C, where all three timepoints are included, the Feb 2026 rate remains the dominant predictor, while the Aug 2024 and Mar 2025 rates flip positive. This means that, conditional on where a practice ends up, having been doing OC earlier is associated with slightly better satisfaction — consistent with a transition effect.

Context: Nearly half of all practices across every tertile had zero OC in August 2024. The big acceleration for the High OC group came after the GPPS fieldwork: their mean rate was 6.6 per 1,000/day in March 2025 but 12.9 by February 2026. The satisfaction scores therefore reflect a much earlier, lower-intensity stage of OC use than the Feb 2026 tertile labels imply.

Method note: All models include list size, IMD 2025, and age as covariates (matching the published model). OC rates converted to monthly submissions per 1,000 patients for comparability. Aug 2024: 22 working days; Mar 2025: 21 working days; Feb 2026: 20 working days. N = 5,741 practices with data at all three timepoints.