Switch teleradiology providers: the 90-day checklist

Switching teleradiology providers fails in predictable places: notice served before the replacement can legally read, a parallel run nobody measured, a cutover with no rollback path, and a closeout that never happened. Every one of those failures is a sequencing error, and sequencing errors are what checklists are for.

This is the 90-day playbook in checklist form: three phases, owners per item, go/no-go gates, and a rollback plan. The companion piece, switching without a coverage gap, analyzes the contract clauses and the compressed 60-day version for a buyer who has already chosen a replacement; this page starts earlier, at the decision itself, and runs through closeout. AstraRad sits on the receiving end of these migrations regularly, and the playbook reflects where they actually strain; if your phase 1 shortlist is still open, a written rate card and licensure attestation reach you within one business day of a request.

When should you switch teleradiology providers at all?

A switch consumes PACS administrator hours, medical staff office cycles, and three months of management attention, so the first checklist item is honest: confirm the problem is the vendor. Three findings justify the cost. Persistent SLA misses documented across multiple monthly reports, after escalation, with the trend still wrong. A structural mismatch no service recovery can fix: preliminary reads where you need finals, no subspecialty coverage for a growing study mix, a minimum your shrinking volume cannot hit. Or an ownership change that reset service quality, pricing, or both.

One rough quarter is usually a renegotiation, and the incumbent's account team knows it; a documented pattern is a migration. If you are still diagnosing which one you have, the provider expectations guide defines what normal service owes you, and the gap between that page and your monthly reports is the evidence either way.

How the 90 days are structured

The phases overlap in one deliberate way: credentialing paperwork starts in phase 1 the moment a replacement is chosen, because licensing and privileging are the critical path of the entire migration and everything else can run beside them.

Days 1 to 30: decide and baseline

Every item has an owner and an exit criterion; treat unchecked boxes as blockers.

• Capture the baseline (imaging director). Pull 90 days of turnaround by tier from your own PACS and RIS timestamps, the addendum rate, clinician complaints, and invoiced cost per study by modality. Exit: a one-page baseline your CFO and chief of staff have both seen. Do this first; it is easiest while incumbent portal access is healthy.
• Read the incumbent contract (administration). Notice period, auto-renewal window, exclusivity, change of control. Calendar the notice deadline itself. The clause-by-clause walkthrough is in the switching guide. Exit: the four clauses summarized in the project file, with dates.
• Shortlist and interrogate replacements (imaging director). Three vendors, one identical written question set: turnaround definitions and trailing compliance, subspecialty roster against your case mix, report tier by hour, complete fees, licensure in your states. The selection guide has the full set and the company comparison the published-claims record. Exit: written answers on file from all three.
• Verify licensure claims (medical staff office). Two named readers per finalist checked against your state board's public lookup. Exit: screenshots in the file.
• Confirm the exclusivity carve-out (administration). If your incumbent contract routes all covered studies to it, get a written pilot exception before any study moves. Exit: the email.
• Sign the pilot scope (administration). Volume split, duration, measures, and pricing for the parallel run. Per-report vendors make this easy: AstraRad prices pilot studies on the same rate card as production, no minimum, so the pilot costs exactly what the studies cost, per the pricing model. Exit: signed scope, start date set.

Days 31 to 60: credential, connect, run parallel

• Start credentialing transfer week one of this phase (medical staff office). For hospitals, confirm whether your bylaws permit credentialing by proxy under 42 CFR 482.22, which turns a committee cycle into a document review; critical access hospitals have the parallel option under 485.616. Request the complete credential file per assigned reader. The mechanics live in the credentialing by proxy guide. Exit: every reader on your account roster credentialed or formally in process with dates.
• Track licensure for any gap states (vendor, verified by you). Where a needed license is pending, the IMLC expedites qualifying physicians in member states; California and other non-members run their own timelines. Exit: no study type in your plan depends on a license that does not exist yet.
• Build and test the image route (PACS administrator). DICOM in both directions or portal upload, priors following their studies, HL7 results landing in the correct RIS queue for every modality, identifier namespace check. The step-by-step is in the PACS integration guide. Exit: a signed test report for every modality, including one study with priors.
• Brief the clinical floor (clinical leadership). ED and ordering providers know the split, the escalation number for the new vendor is posted at the console, and technologists know which studies route where. Exit: one page, posted, acknowledged.
• Run the parallel for two to four weeks (operations). One owner per study, split by study type, hour block, or facility. Measure both vendors identically: turnaround by tier, escalated-case quality, time to reach a radiologist, addendum rate. Test the 1 a.m. escalation call once, deliberately.
• Score against the go/no-go gates (imaging director). Gates written before the run started. Exit: every gate green, or the run extended with a dated re-review.

Days 61 to 90: cut over, then actually finish

• Serve notice (administration). Timed so the notice period expires just after planned cutover completion. This is the irreversible step; everything above happened quietly while the incumbent was still reading.
• Cut over one tier at a time (operations). Routine first, then urgent, then STAT last, once the escalation path has proven itself on live cases. Audit sent-versus-ordered counts daily during the moves.
• Export before access ends (operations). Portal-only reports, QA records, discrepancy logs, compliance history. Incumbent portals close fast after termination, and re-requesting records later is slow at best.
• Reconcile the final invoice (finance). Line-by-line against your own study counts for the stub period, minimums and true-ups included; the patterns to check are in the hidden fees guide.
• Confirm records retention in writing (compliance). What the departing vendor retains, for how long, under whose obligation, and the destruction certification at the end of it.
• Run the 30-day lookback (imaging director). New vendor's first full month against the phase 1 baseline, on the same four measures. Exit: the lookback filed and shared with the stakeholders who approved the switch. This is the item that makes the next renewal negotiation easy, whichever vendor it is with.

The gates and the rollback plan

Write the go/no-go gates during phase 1, while nobody is emotionally invested in the answer. A workable set: 95 percent of parallel-run studies inside their stated tier, measured from last-image arrival to signature; zero unresolved critical-findings communication failures; credentialing complete for the full account roster; results confirmed in the correct RIS queue for every modality; and the escalation drill answered inside the vendor's stated response time. Publish the gates to both vendors. The incumbent behaves better during a measured trial, and the challenger knows exactly what it is being scored on; a vendor that resists written gates has failed one.

The rollback plan is one paragraph and it should exist in writing before the first tier moves: the incumbent route stays configured and tested until closeout, notice timing preserves incumbent coverage through cutover, and any failed gate after a tier moves reverts that tier within one day while the failure is worked. Rollbacks executed against a written paragraph are boring; rollbacks improvised at 4 p.m. on a Friday are how coverage gaps happen.

Three failures this sequencing prevents

The notice-first failure. An imaging director, angry after a bad month, serves 60-day notice on day one and then starts shopping. Selection takes three weeks, credentialing the incoming readers takes six more, and the incumbent's notice period expires with the replacement still unable to legally sign reports in the state. The facility bridges the gap with expensive locum coverage or, worse, with unread studies. In the playbook above, notice sits at day 61 for exactly this reason: it is the only step that cannot be undone, so every slow, reversible step runs first, quietly, while the incumbent keeps reading.

The unmeasured parallel run. Both vendors go live, the month passes, everyone has impressions and nobody has numbers. The loudest anecdote wins: one missed STAT read from the challenger, remembered vividly, outweighs four weeks of quietly better routine turnaround, or the reverse. Cutover proceeds on politics. The fix costs one spreadsheet: four measures, both vendors, same definitions, reviewed against gates that were written before the run started. Measured parallel runs also change vendor behavior mid-trial, because both sides know the score is being kept.

The closeout that never happened. The new service is live, the team is tired, and the last four checklist items quietly die. Eight months later a payer audit needs a 2024 report that lived only in the departed vendor's portal, access to which ended with the contract; the final invoice's volume true-up was never reconciled and quietly overcharged; nobody can prove the switch improved anything because the baseline was never compared against month one. Every item in the closeout list is ten times cheaper in week twelve than in month twelve, and the 30-day lookback is what turns the whole migration from an act of frustration into a documented, repeatable capability your facility now owns.

The pattern across all three: the expensive mistakes are calendar mistakes. Vendor quality matters enormously at selection time, and by phase 2 the risk has moved entirely into your own sequencing, which is the part no vendor can do for you.

What the receiving vendor owes you during all this

The migration workload is real, and a replacement vendor should carry its share: a named onboarding owner with a dated plan, credential files delivered complete instead of dribbled, integration engineers who have done your PACS before, pilot pricing identical to production pricing, and written answers fast enough to keep your checklist moving. AstraRad's standing commitments to a switching buyer: a written rate card within one business day of a request, licensure attestation for the radiologists who would read your studies, final signed reports from the first study, and tiers of STAT under 4 hours, with 2-hour and 1-hour STAT available on request at a higher multiplier, and routine 24 to 48 hours, with the measurement method published on the SLA page. The first signed report lands within 10 business days of countersignature, which means a phase 2 parallel run can start inside two weeks of a signed pilot scope.

Ninety days sounds long until it is week seven and the checklist is the only thing keeping four departments synchronized. Print it, assign the owners in the first meeting, and let the gates make the hard calls. A migration run this way ends with something better than a new vendor: a baseline, a measured comparison, and a closeout file that makes every future renewal negotiation start from evidence instead of memory.