InnovaRx Innova Health Strategies

How to use InnovaRx

A short guide to each page — what it is for, how to work it, and where its edges are. Every page carries the same guide behind its How to use this page panel.

Start here

Four ways in, depending on what you are holding.

  1. Have the words from a note — "diabetic neuropathy"? Use Index a code, which walks the alphabetic index the way the book does.
  2. Have a code already? Type it in the search bar at the top; it goes straight to the code page.
  3. Want to read a section of the index in order? Use Browse.
  4. Want to know what a medication implies about documentation? Use Meds.
/
jump to the search bar from anywhere

Index a code

The alphabetic index, walked one level at a time — the same route the printed book takes.

  1. Type the main term from the documentation — the condition, not the body part. "Diabetes", not "foot".
  2. Press Tab to commit that term. It stays on screen as a crumb, because the route is part of the answer.
  3. Type the next word to filter that term's subterms. "w" offers "with"; the list only ever shows subterms of where you are standing.
  4. Keep going until a subterm carries a code. Press Enter to open it.
  5. Click any crumb to go back to that level, so a wrong turn costs one click.
Tab
accept the highlighted term and descend
↑ ↓
move through the subterms
Enter
open the code
⌫
on an empty box, climb back a level

What this page does not do

  • The index sends you to a category; the tabular is what confirms it. Where the index stops short — laterality, staging, a 7th character — the page says so rather than handing you an incomplete code.
  • Subterms appear in the index's own order, not ours. That is why "with" sits above "without".

Browse the index

The alphabetic index read in order, A to Z, rather than searched.

  1. Pick a letter, then scan the main terms under it.
  2. Open a main term to see its subterms nested underneath, indented as in the book.
  3. Use this when you want to see what sits *near* a term — the neighbours are often the reason the index is worth reading rather than querying.

A code page

Everything the published sources say about one code, on one page.

  1. Read the Excludes1 block first — it is the one that stops you coding this at all alongside another code.
  2. Excludes2, Code first, Use additional and Includes follow. A note inherited from a parent category is labelled with the category it came from, so you can see why it applies.
  3. Official Guidelines sit below. "Guidelines naming this code" cite the code or a range containing it; "Chapter guidance" is the general guidance for its chapter and is the weaker signal.
  4. Guideline text is quoted verbatim with its section number and page, so you can check it against the published document.
  5. Risk-model chips show which published CMS models map this code.

What this page does not do

  • Guidance is quoted, never paraphrased — a paraphrase of a coding rule is a different rule.
  • Medication candidates, where shown, are matched by condition *name* and are offered as candidates for you to judge. They are not "the drugs for this code".

RAF calculator

What the published CMS-HCC models do with a set of codes.

  1. Add the codes you want to evaluate, then set the demographic factors the model takes. Dots are optional — a paste straight out of a claims extract works.
  2. The result shows each code that maps to an HCC, the coefficients applied, and the total the published model produces.
  3. Two lists come back separately at the bottom. A code that maps to no HCC is ordinary — most diagnoses do not risk-adjust. A string that is not a code in that year at all is a typo, or a sign the code set year does not match the dates of service.
  4. Use it to check your understanding of a model, or to see why a set of codes scores the way it does.

What this page does not do

  • This is arithmetic on the published model, nothing more. It is not advice about which codes to report.
  • A code belongs on a claim when the documentation supports it. That question is answered in the chart, and no calculator can answer it.

Batch risk scoring

The same model arithmetic, run across a roster instead of one member.

  1. Paste the roster — a header row in any order, or no header at all. Unrecognised columns after the named ones are read as more codes, which is what a dx1…dx12 export looks like.
  2. Pick one model, one payment year and a default segment. A segment column in the roster overrides the default row by row.
  3. Every row comes back with its score, the HCCs behind it, and what the hierarchy suppressed. Rows that could not be scored say why instead of showing a zero.
  4. Download the CSV for a spreadsheet. It is the same table, in the order you submitted it.

What this page does not do

  • Nothing is stored. The roster is scored in memory and rendered back; no table, file or log line keeps it.
  • The member id is passed through and never interpreted — use your own internal key, not a name, an MBI or a date of birth.
  • No ranking and no deltas. Rows come back in submitted order, and nothing here suggests what would change a score.
  • A row whose factor has no coefficient is a floor, not an answer, and is marked as one. The mean includes it, so the mean is a floor too.

HCC overlap check

What a set of codes actually yields once the model collapses the duplicates.

  1. Paste the codes — commas, spaces or new lines, with or without the decimal point.
  2. Pick the code set for the date of service. A code only maps in the years it existed.
  3. Each model is shown separately, because which one applies depends on the payment year and the plan.
  4. Two codes on one HCC count once. An HCC struck through is outranked by a higher one in the same family and contributes nothing.
  5. The link at the bottom lays the same codes across every loaded year, which is the view for an audit that spans years.

What this page does not do

  • No score. This reports which HCCs a code set yields, not what they are worth — scoring needs coefficients, and for the historical years those are not loaded.
  • It is not advice about which codes to report. A code belongs on a claim when the documentation supports it.

Code history

Whether a code was valid on the date of service, and what it mapped to under the model that paid that year.

  1. Paste the codes, and give the date of service if you have one.
  2. The code set turns over on 1 October, so a date in October, November or December belongs to the NEXT fiscal year — 4 November 2021 is FY2022. The page does that conversion for you.
  3. The verdict block answers the date directly. The table below it shows every year loaded, so you can see when a code arrived, when it went, and when its wording changed.
  4. The change log flags what matters in a review: new, deleted, reworded, billable status, and any move between HCCs within the same model.

What this page does not do

  • A year with no model files loaded reads "no model loaded", never "maps to nothing". Those are different facts and only one of them is about the code.
  • HCC movement is compared within a model. The V24 to V28 changeover is a change of model, not a change to the code, and is not reported as one.
  • It reports what the published files say. Whether a code belonged on that claim is answered by the documentation.

Medications

What NLM's MED-RT says about a drug — what it is indicated for, and what it is contraindicated with.

  1. Search a drug name to see its conditions, grouped as indicated for, indicated to prevent, contraindicated with, and may cause.
  2. Search a condition name to go the other way.
  3. Use it when a note mentions a medication and the question is what that implies about the documentation.

What this page does not do

  • MED-RT keys conditions by MeSH descriptor, not ICD-10. The two carve the world differently, so a condition here is a name to reason from, not a code.
  • A medication is never evidence of a diagnosis. It is a reason to go and look at what the chart actually says.

NPI lookup

Who a provider number belongs to, straight from the federal registry.

  1. Type what you have into the one box — a number, "Smith, John FL", a practice name. It is parsed into the fields below, where you can correct it.
  2. You do not need all of it. A last name alone works; so does a first name with a state.
  3. If nothing matches everything you gave, the least reliable field is set aside and it searches again — the postcode first, then the city, the taxonomy, the first name, and the state last. Whatever finally ran is named above the results, so a broadened match is never mistaken for an exact one.
  4. First names match nicknames, so Bob finds Robert. Apostrophes are dropped because the registry does not store them.
  5. Each result carries every taxonomy with its state licence, every address on file, and any other number the provider is known by — a state Medicaid id, a payer-assigned number.
  6. The NPI check digit is verified before the registry is called, so a mistyped number is caught here rather than coming back empty.

What this page does not do

  • This is NPPES, the public federal registry, read live. It is not a credentialing check, a sanctions check, or proof of participation in any plan.
  • Registry records are self-reported by the provider and can be out of date. The date NPPES last updated each record is shown for that reason.
  • A city or state matches ANY address on the record, so a provider can turn up under a state they do not practise in. The address that matched your search is shown first for that reason.
  • NPPES never reports a total. A result list marked "20+" means there were more, not that there were exactly twenty.

Bookmarks

The codes you keep coming back to, kept against the year they were saved.

  1. Star a code from its page to save it. Starring again removes it.
  2. A note on a bookmark is yours — a reminder about a payer rule, a query you keep having to write, a distinction you keep having to look up.
  3. The list checks every bookmark against the code set you are viewing, so a code that was deleted or revised in a later year says so rather than quietly disappearing.

What this page does not do

  • Codes only. A bookmark holds a code, your own note and nothing else — never anything from a chart.
  • Bookmarks belong to you, not to your account. A colleague on the same team does not see them.

Your account

Plan, billing and — for a practice — the people on it.

  1. The free tier allows a set number of code pages a day. Past it the page still renders in full; you are asked, not blocked.
  2. Manage payment details, invoices and cancellation through the billing portal.
  3. For seats across a practice, get in touch — team plans are quoted rather than self-serve.

The browser extension

The same index walk, in a popup, for Chrome and Edge.

  1. Open it with the toolbar button or Ctrl+Shift+K.
  2. It walks the index exactly as the site does — type, Tab to descend, Enter to open a code.
  3. The code appears below with its Excludes, instructional notes and the Official Guidelines that name it, each foldable.

What it does not do

  • It reads nothing on the page you are on. There is deliberately no content script, so the extension never asks for access to whatever chart or EHR you have open — text reaches it only when you send a selection to it yourself.