Park Column — Hygiene Visits / Hygienist-Day v2

SGA West · 56 live offices · FTE-normalized (visits ÷ hygienist days worked), from the PBI · matched against a non-Park-Column baseline · pulled 2026-06-10
Where to push next — ranked by opportunity, not delta
A priority score per GP office: headroom to the high-performing baseline ceiling (room to grow) + momentum (declining offices float up) + an actionability flag (low form usage = a concrete lever), all weighted by hygienist FTE so big teams matter more. Click any row for the full record.

Does the edge survive level-matching?

Each Park Column GP office is paired with its nearest non-Park-Column twin by absolute YTD’25 baseline (the level the cohorts barely share). We then compare each office’s YTD change to its twin’s — a per-pair difference-in-differences — and aggregate. The raw headline compares a +2.5pp edge across unmatched populations; this asks whether it holds when each office faces a same-level peer.

Baseline-level overlap — the two populations barely sit on top of each other

35791113
Park Column GP offices (YTD’25 level) Non-PC control offices (YTD’25 level)
Each tick = one office’s absolute visits-per-hygienist-day baseline. Control clusters higher and stretches to a 13.2 perio-heavy outlier; matching pulls each PC office to a same-level peer (median match distance < 0.1 VpD) so the comparison is fair.

Matched pairs — each office vs its closest twin

Sorted by per-pair diff-in-diff (office YTD% − twin YTD%). Green = the office out-improved its same-level peer. Click a row to open the office record.
Park Column office Base ’25 Office YTDΔ Matched twin (non-PC) Match gap Twin YTDΔ Diff-in-diff

Usage vs improvement — the most important caveat, as evidence

Every dot is one of the 47 GP offices: form submissions (usage proxy) on x, YTD change in visits/hygienist-day on y. If the column drove the gain, the cloud would slope up. The fitted line is nearly flat / slightly negative (r = −0.135) — high-usage offices did not improve more. The horizontal line is the control’s YTD change.
Dormant (0 submissions) Occasional (1–9) Regular (10+) OLS trendline

By reporting intensity — no dose-response

If the column drove throughput, regular reporters should clearly beat the control. They don’t — dormant offices improved most.
CohortOffices YTD ’25YTD ’26YTD Δ May ’25May ’26May Δ

On the FTE-correct metric — hygiene visits per hygienist per day — Park Column GP offices improved +1.9% YTD while comparable non-Park-Column offices slipped −0.6%: a +2.5pp raw edge. New in v2: pair each office with a same-level twin and the edge holds and widens to +4.5pp (+3.3pp FTE-weighted). The catch is unchanged — offices that used the column most did not improve more.

What this tells us

  • Normalize by hygienist, and a modest edge appears. Visits per hygienist-day strips out headcount — an office that added a 4th hygienist no longer looks busier just for it. On that basis Park Column GP rose +1.9% YTD vs the control’s −0.6%, a +2.5pp gap. May was roughly flat (−0.4pp).
  • v2 answer to the “unmatched populations” objection. The standing weakness was that the control runs ~7.9 vs ~5.9 visits/hygienist-day — a different population. Pairing each of the 47 offices with its nearest-level non-PC twin (median match distance < 0.1 VpD) gives a clean diff-in-differences of +4.5pp, FTE-weighted +3.3pp. 30 of 47 offices beat their same-level peer. The edge is not an artifact of the level gap.
  • But usage still doesn’t explain it. If the column caused the lift, heavy reporters should lead. They don’t: regular +1.2% YTD, occasional +0.2%, dormant +5.4% (highest). The fitted slope of submissions vs improvement is slightly negative, correlation −0.135. So the edge is real but not cleanly the column at work.
  • The move is FTE, not raw volume. The biggest per-FTE gainers got there by shedding hygienist-days, not adding visits — Wagner Dental’s +41% came from visits +16% while hygienist-days fell 18%. The metric is doing exactly what it should: rewarding throughput per hygienist, not headcount.
  • Net read. Cautiously positive — and more robust than v1 now that the edge survives matching — but still short of a clean causal win. Confirm with consistent daily use, then re-measure; and get appointment-level (CDT) data to isolate the slots the column actually recovered.

The fuller picture

What the Park Column is meant to do. When an appointment isn’t confirmed ~24 hours out, the office moves it into a “park” column and backfills the freed slot — the goal is to stop losing hygiene chair time to broken appointments. If it works, a rollout office should get more completed hygiene visits out of each hygienist-day than an office without the column.

Why per-hygienist, and why a matched baseline. Dividing by office operating days inflates the instant an office adds a hygienist. The correct normalization — the one SGA East uses — is visits per hygienist-day: completed hygiene visits ÷ hygienist days worked. We compared the 47 Park Column GP offices against the non-PC West offices over the same periods, and now also office-by-office against a same-level twin so the structural level gap can’t carry the result.

What we found. Per hygienist-day, Park Column GP went 5.83 → 5.94 YTD (+1.9%) while the control went 7.93 → 7.88 (−0.6%) — a +2.5pp raw edge. The matched-pairs cut tightens and strengthens this to +4.5pp (mean office +3.9% vs mean twin −0.6%), FTE-weighted +3.3pp, with 30 of 47 offices beating their peer. But splitting the rollout offices by reporting intensity shows no dose-response: regular +1.2%, occasional +0.2%, dormant +5.4%, correlation −0.135.

How to read it honestly. The YTD edge is genuine and now survives level-matching — the strongest version of the “it worked” case the data supports. What it still can’t do is pin the gain on the column: usage doesn’t track improvement. Without CDT-level appointment data we can’t isolate the specific appointments the column recovered.

Bottom line. On the FTE-correct metric, Park Column offices edged a matched baseline by +4.5pp YTD — a more robust read than v1 — but the absent usage dose-response keeps it short of proof. The Focus Priority tab turns this into the next move: where to drive daily use and audit hygiene schedules.

Office ROD Subs May ’25 May ’26 May Δ YTD ’25 YTD ’26 YTD Δ
Click any row for the full office record + the “why it moved” FTE decomposition. Δ = % change in hygiene visits per hygienist-day (FTE).

Hygiene VpD movement by ROD

Offices with PBI data (GP + perio). Navy bars = YTD ’26; gold line = YTD ’25.
RODOffices May ’25May ’26May Δ YTD ’25YTD ’26YTD Δ
Heidy Riall’s portfolio is entirely periodontics — the Park Column is not actively run there.

The question

Did hygiene visits per day actually improve at the 56 SGA West offices that rolled out the Park Column — measured from the PBI, because the offices won’t keep the manual daily form (22% form compliance)? This is the SGA West parallel to the SGA East risk-score readout.

Metric definition (FTE-normalized)

Hygiene Visits per Hygienist-Day = completed hygiene visits ÷ hygienist days worked, per office per period. Completed hygiene visits = [Completed Visits] where User Type = "HYG"; hygienist days worked = [Work Days] where User Type = "HYG" (the FTE denominator). Cohort numbers are volume-weighted (Σ visits ÷ Σ hygienist-days).

Matched-control method (new in v2)

The standing weakness of the raw comparison was that the control runs a structurally higher level (~7.9 vs ~5.9 VpD), so the cohorts aren’t matched. For each of the 47 GP offices we pick its nearest non-Park-Column twin by absolute YTD’25 baseline (minimizing |office.vpd_YTD25 − control.vpd_YTD25|) from the 23 control offices with usable data, then compute a per-pair diff-in-difference (office.vpd_YTD_pct − twin.vpd_YTD_pct) and aggregate. Median match distance is < 0.1 VpD, so each office faces a genuinely same-level peer. Result: matched edge +4.5pp simple, +3.3pp FTE-weighted, 30/47 offices beating their twin. The control set carries no ROD field, so the optional same-ROD constraint could not be applied; matching is on level alone.

Usage dose-response method

We regress each office’s YTD change on its form-submission count (usage proxy) across the 47 GP offices: OLS slope −0.20, intercept 6.1, Pearson r = −0.135. Tiers: Regular reporters (10+ subs, n=22), Occasional (1–9, n=14), Dormant (0, n=11). A real column effect should show most in heavy users; the slope is slightly negative instead.

Priority score method

Per GP office: 0.45·headroom + 0.40·momentum + 0.15·actionable, each min-max normalized across the 47, then multiplied by √(hygienist provider-days) so high-FTE offices weigh more without fully dominating. Headroom = control YTD’26 ceiling (7.9) − office YTD’26; momentum = −YTDΔ (decliners float up); actionable = 1 if subs < 10. The recommendation line is templated from each office’s own fields.

Periods & source

May 2026 vs May 2025, and YTD Jan–May 2026 vs 2025. Park Column went live Oct/Nov 2025. Gen4 / SGA West Power BI dataset, queried via Service Principal; the 56 live offices matched by practice name to the PBI location dimension.

Coverage & exclusions

  • 47 GP offices — the core Park Column cohort; the real test.
  • 6 perio offices (Heidy Riall’s portfolio) — shown separately; perio teams aren’t actively using the Park Column.
  • 3 offices excluded — no usable PBI hygiene data: MCP Salinas, DDM Bloomington, ACR LDC.

Read it carefully

Hygiene VpD moves for reasons beyond the Park Column (staffing, demand, schedule template). This shows whether throughput moved in the right direction — it is not a clean attribution of the column alone. Without CDT-level appointment data we can’t yet prove the office moved the right appointments.

All figures computed client-side from the same PBI aggregate bundle as v1 (no patient-level data). Public aggregates only.