Healthcare data analytics reads the EHR, the practice system, the lab, and payer remittance as one patient and one visit. One view matters because a stay, a denial, and a lab delay can belong to the same encounter, and each system describes it differently.
Why the systems disagree
Four sources cover a single stay, and none of them was built to answer the others’ questions.
The EHR holds the clinical record: problems, orders, notes, and disposition. A visit there is a documented encounter.
The practice management system holds the schedule, registration, charges, payments, and adjustments. A visit there is an appointment that was kept, or a claim that was billed.
The laboratory information system holds specimens, results, and critical flags. The unit of work is often a specimen, not a patient day.
Payer remittance holds the allowed amount, the paid amount, denial codes, and patient responsibility. A visit there is a claim, or a line on a claim, and it arrives days or weeks after the encounter.
They disagree in four places.
Identifiers. The medical record number, the billing account, and the subscriber ID are different keys for one person. Matching on name joins the wrong patients and misses a registration under another spelling.
Encounter identifiers. The visit number, the appointment number, and the claim number are not the same field. Without a map, today’s visit count is three different numbers.
Provider names. The attending on the note, the rendering provider on the claim, and the ordering provider on the lab slip are often different strings for one person. A report by provider splits that person into three rows until the strings are mapped.
Dates, and the word visit. Arrival, discharge, collection, result release, charge post, and adjudication are six dates. A visit can mean a documented encounter, a billed claim, or a specimen. Until the page names the date and the meaning it uses, the systems look like they contradict each other. They are counting different things.
Four sources, four meanings of a visit
What each system holds, and what it counts.
| Source | What it holds | What a visit means |
|---|---|---|
| EHR | Problems, orders, notes, disposition | A documented encounter |
| PMS | Schedule, registration, charges, payments | An appointment or a claim |
| LIS | Specimens, results, critical flags | A specimen |
| Payer remittance | Allowed, paid, denied amounts | A claim or a claim line |
What one view takes
One view comes down to three decisions, made once. The pipeline that carries them is data engineering work. The page a leader opens is the business intelligence view.
A shared patient key and a shared encounter key. Map the medical record number to the billing account, and the visit number to the claim. Do the mapping in a table someone owns, and keep the failed matches on a list. A row that does not match is a registration problem, not a blank on the dashboard.
One definition per metric. Write the formula, the date it uses, and what is in or out. If finance and operations both say “visits,” they use that sentence.
A refresh schedule tied to the decision. Bed status and emergency department flow need to move through the day. Days in accounts receivable and staffing ratios can wait for a nightly load.
What one view takes
Key, definition, refresh.
- 1
Shared keys
Patient and encounter, mapped, not matched on a name.
- 2
One definition
Each metric written once and used in every report.
- 3
Refresh schedule
Flow through the day. Revenue cycle nightly.
KPIs by function
Use the formulas below as the definitions. They are ratios, not targets. A target is a decision the hospital makes. The page should not invent one.
Patient flow
Average length of stay = total inpatient days / discharges
Bed occupancy rate = occupied beds / available beds
ED door-to-provider time = average minutes from arrival to first provider contact
Left-without-being-seen rate = patients who left before a provider / ED arrivals
Appointment no-show rate = missed appointments / scheduled appointments
Length of stay matters when discharge is what frees the bed. Occupancy matters in the morning huddle, when the question is whether elective cases can proceed. Door-to-provider time and the left-without-being-seen rate matter while the waiting room is still filling, not in next month’s packet. The no-show rate matters when tomorrow’s schedule can still be changed.
Revenue cycle
Days in A/R = total A/R / (gross charges for the last N days / N)
Clean claim rate = claims accepted on first submission / claims submitted
Denial rate = denied claims / claims submitted
Net collection rate = payments / (charges - contractual adjustments)
N is the number of days in the window, and it stays the same every month. Days in accounts receivable matters when cash slows and the cause might be coding, denials, or a payer. Clean claim rate matters before a denial exists, and denial rate matters when choosing which reason to work. Net collection rate matters at the monthly close, and it is not cash posted this week.
Lab
Turnaround time = result time - collection time
Critical result notification time = time from a critical flag to notice reaching the ordering clinician
Calculate turnaround time by test priority. A stat potassium and a routine panel are not one average. Turnaround time matters when a unit is waiting on a result to discharge or to treat. Notification time matters when the result is critical. The clock stops when the ordering clinician is reached, not when the instrument finishes.
Staffing
Nursing hours per patient day = nursing hours / patient days
Overtime percentage = overtime hours / total nursing hours
Contract labor percentage = contract hours / total nursing hours
Nursing hours per patient day matters when census and the roster are being set for the next shift. Overtime percentage and contract labor percentage matter when the question is the cost of covering the census, not the headcount on the schedule. Use paid hours from the same source every time. A schedule that was never worked will make the ratio look better than the shift was.
Aligning EHR analytics and payer data
Clinical counts and billing counts split in three ways, and the page should show the split instead of hiding it.
Denials. The EHR can show that the service was documented. The remittance can show that it was not paid, with a reason code. Joined on the encounter key, that is one visit with two outcomes. A page that starts from only one of those sources cannot see the gap.
Coding. The problem list and the diagnosis on the claim are written at different times, often by different people. If the page uses the claim diagnosis for a clinical question, it is answering the coded bill, not the record. Keep both, and label which one the chart uses.
Timing. Charges post after the visit. Adjudication is later still. A census for Tuesday and a cash report for Tuesday are not the same set of encounters. Reconcile on the encounter date. Show the lag as its own column: days from discharge to charge post, and days from submission to remittance. The lag explains why this morning’s clinical total and this morning’s paid total do not match. They are not supposed to, yet.
On one page, use the same encounter key and the same date definition. Put the clinical count, the charged count, the paid count, and the denied count in one row. Filter by service line and by payer. The work list is the rows where those counts disagree.
What the operations dashboard shows
These numbers are made up to show the layout. Occupancy is 80 percent. Door-to-provider time is 30 minutes. Length of stay is 4 days.
The left-without-being-seen rate is 2 percent. The no-show rate is 10 percent. The denial rate is 8 percent.
The top of the page is the huddle, in four cards: bed occupancy, door-to-provider time, length of stay, and left without being seen. Each card shows today against the same day last week. A leader who has two minutes stops here.
Below the cards:
- Emergency department arrivals by hour, as a line, so a backlog is visible before the waiting room is full.
- Length of stay by unit, as a bar, so the bed meeting knows which unit is holding patients.
- No-show rate by clinic, for the clinics whose schedule can still change.
- Denials by reason for the open work queue, using the encounter key so each reason ties back to a visit.
Leave these off this page: every lab result, every claim line, a score with no definition, and a clinician leaderboard. The result belongs on the lab bench. The claim line belongs with the person posting it. A ranking turns the huddle into a contest.
Staffing ratios can sit on a second page for the person building tomorrow’s roster. They do not need to lead this view.
Period, site, and unit filters sit above the cards and move every chart together. A chart that ignores the filter does not belong here.
An operations dashboard
Round numbers, made up, to show the layout.
Leave these off
What does not belong on the operations page.
- Every lab result
- Every claim line
- A score with no definition
- A clinician leaderboard
PHI in a BI environment
The model behind the operations page holds patient data, even when the page only shows rates. Treat the model the way the source system is treated.
Give each role the minimum it needs for its decision. The huddle needs unit, hour, and a rate. It does not need a named diagnosis, a medical record number, or the note. A row that is not required for the decision stays in the source system.
Power BI, Tableau, and Qlik all support row-level security (Qlik calls it section access). Use it so a user sees the sites and units their role allows. A filter on the page is not a control. The restriction has to live in the model, or a user who can reach the table can see past the filter.
A view for a wide audience should be aggregated, or stripped of direct identifiers, before it is published. Occupancy for a unit is a different report from a list of who is in the beds. Publish the list only to the role that staffs those beds.
Log who opened which report, and review that log on the schedule the compliance officer sets. Any vendor that stores or processes this data needs a business associate agreement. That agreement is a contract the compliance officer accepts or refuses. It is not a box in the dashboard tool, and it is not a label a product can put on itself.
The compliance officer makes the final call on what leaves the source system and who may see it.
What real-time means in practice
Real-time means the number is current enough for the decision in front of you. It does not mean every edit in the EHR appears on the page within a second.
Near-real-time is for emergency department flow and bed status: door-to-provider time, arrivals by hour, and occupancy, while the shift is still on. Refresh every few minutes. If the feed only arrives hourly, print that time on the card rather than calling the number live.
Daily is enough for staffing ratios and for lab turnaround as a trend. The roster for tomorrow is set from yesterday’s worked hours and today’s census. A turnaround-time average does not need to move every minute to show that stat tests are missing their window.
Monthly is the right grain for most of the revenue cycle. Denial rate, clean claim rate, net collection rate, and days in accounts receivable describe a close, not a huddle.
Put the as-of time on every card. A door-to-provider number from 6 a.m. is not the number for the 4 p.m. surge.
Why visualization matters in healthcare
The monthly total hides the hour. That is the practical case for a chart in this work.
Arrivals drawn by hour show the surge before the average does. A day that looks normal in the daily ED total can still have a two-hour pile-up at 5 p.m. The line makes the pile-up obvious. The total does not.
Length of stay drawn by unit shows which unit is holding beds. A hospital-wide average of 4 days can be three units at 3 days and one unit at 7. The bar shows the comparison. The average hides it.
Denials drawn by reason, beside the service line, show whether the unpaid work is a coding pattern or a single payer. In the example, a denial rate of 8 percent does not say what to fix. The breakdown does. A chart that uses a different meaning of visit than the card above it only gives the meeting a second number to reconcile.
01 FAQ
Frequently asked questions
Why does data visualization matter in healthcare?
It matters when the pattern is in the shape of the day, not in a monthly total. Arrivals by hour show a backlog forming. Length of stay by unit shows which unit is holding beds. A denial table next to the service line shows where the clinical record and the payment split.
What are the benefits of data visualization in healthcare?
A leader can compare units on the same definition, see which hour the emergency department fills, and see which denial reason is growing. The benefit is a decision in the meeting, not a longer report. The chart only helps if the metric underneath it has one definition.
What is a 360 view of patient data?
A 360 view is one patient and one encounter across the EHR, the practice management system, the lab, and payer remittance. It depends on a shared patient key and a shared encounter key, not on matching names. Clinical, scheduling, lab, and payment facts then describe the same visit.
How do you report on healthcare data efficiently?
Stop rebuilding the same metric inside each source system. Write one definition, refresh on a schedule that matches the decision, and put the huddle numbers on one page. Flow metrics refresh through the day. Revenue cycle and staffing can refresh nightly.
How do you align EHR analytics with payer dashboards?
Join them on the encounter key and use one date, usually the encounter date, not the day the cash posted. Show the clinical count beside the paid count and the denied count. The gap is the work: a service documented and not paid, or paid under a code that does not match the record.
What should a healthcare data strategy cover?
The shared patient and encounter keys, one definition per metric, and a refresh schedule by decision. It also covers who may see which rows, what is aggregated before a wide audience sees it, and which vendor agreements apply when PHI moves. The compliance officer decides the access rules.