16.4 Comparative Analytics and Presenting Recommendations
Key Takeaways
- Task A.11 is employ comparative analytics—indicators and benchmarks—to see relative performance, not to decorate a slide with an unmatched peer.
- Internal trend, peer group, and target are different comparisons. A critical-access hospital is not an unadjusted academic medical center, and HIMSS does not publish a CPHIMS-official benchmark set.
- Task A.19 is present interpretations and recommendations of data analyses to decision makers: facts, interpretation, residual, and a real choice—not a 40-page export.
- A recommendation names what a yes displaces and who owns the decision. Decision makers decide; analysts do not smuggle the choice in a color scale.
- Causality is earned. A benchmark gap is a question until design, risk, and residual data have been examined.
16.4 Comparative Analytics and Presenting Recommendations
Quick Answer: Task A.11 is employ comparative analytics (for example indicators and benchmarks). Task A.19 is present interpretations and recommendations of data analyses to decision makers. A red bar against the wrong peer is not analysis. A 40-page appendix is not a recommendation.
Sections 16.1 and 16.2 asked whether we performed and whether users were served. A.11 asks compared with what. A.19 asks so what, said to whom, with a choice attached. Together they close the management loop that started in chapter 15: measure aims, then help someone decide. Domain 4 is still 25% of the 100-scored-item, two-hour exam. Cognitive level is usually application or analysis, not recall of a vendor’s marketing quartile.
HIMSS does not publish a CPHIMS-official benchmark book, Digital Health score you must cite, or slide template. Use the comparison the organization can defend.
A.11 What “comparative” actually compares
| Comparison | Honest use | Dishonest use |
|---|---|---|
| Self over time | Same definition, same grain, marked process changes | Resetting the definition until the line goes up |
| Target / goal | The numbered aim from the plan or the SLA | A target invented after the result is known |
| Internal peer | Like units, after case mix or volume context | The worst unit as the only baseline so everyone looks fine |
| External benchmark | Documented peer group, period, and specification | An academic flagship’s unadjusted mortality next to a critical-access hospital |
Employing comparative analytics means you choose the comparison on purpose:
- Name the indicator and its specification (numerator, denominator, exclusions, lag). “Sepsis performance” is not an indicator.
- Name the comparator and why it is fair. Bed size, teaching status, payer mix, rurality, EHR tenure, and which encounters count all change the story.
- Risk-adjust or stratify when the outcome needs it. If you cannot adjust, show crude rates and the limitation. Do not pretend.
- Separate IT-enabling measures from care outcomes. Portal logins are not a mortality benchmark. Interface success is not a length-of-stay benchmark.
- Mind the lag and the source. Claims, EHR, registry, and survey files are not interchangeable without a residual note (chapter 15’s honesty rule still applies).
- Treat a gap as a question. Being in the worst quartile of scan compliance can mean devices, workflow, staffing, or a broken denominator. A.11 finds the gap; it does not automatically buy a new module.
External numbers show up as CMS and other public-quality programs, payer scorecards, safety organizations, state hospital associations, and subscription peer services. Vendor “average customer” charts are marketing until you can see the spec. Do not cite a folklore HIMSS stage or a made-up national HIT happiness index as if the exam handed you a table.
In a health system, comparative work is often hospital-to-hospital or clinic-to-clinic with a shared definition. In an FQHC, the honest peer is other community health centers, not an academic transplant program. In a payer or public-health shop, the grain may be member-month or catchment—not discharge.
A.19 Present interpretations—not a tour of the warehouse
The handbook task is present interpretations and recommendations of data analyses to decision makers. That is a communication and judgment task:
- Decision maker means the person who can allocate capital, stop a project, change a policy, or accept a residual—board, operating executive, medical staff leader, CFO, or a governed committee. Sending the same 40-page pack to everyone is not tailoring.
- Interpretation is what the analysis means, with confidence and residual. “Scan compliance is 71% versus an 85% internal target and versus 82% in the peer group of like medical-surgical units; night shift and two cart-poor units drive the gap; device downtime is in the ticket themes.” That is an interpretation.
- Recommendation is a choice with consequences: stabilize devices and restaff the swap service before buying another scanning product; or accept the residual and say so. Name what a yes displaces.
A usable A.19 brief, regardless of slide tool:
- Decision needed, owner, and date.
- Indicator and comparator (A.11), in one sentence.
- Finding, including who is in the denominator.
- Interpretation: likely drivers you actually checked, not a novel.
- Residual and data limits.
- Options (at least two real ones), cost/capacity, and what each displaces.
- Ask. Then stop talking.
Do not bury the ask on slide 27. Do not smuggle a vendor award inside a heat map. Do not claim the benchmark caused the outcome. Do not hide uncertainty to look decisive; decision makers can handle a confidence range. They cannot handle a surprise after they voted.
Match the room. The board needs organizational grain and residual honesty (section 15.1). Medical staff need clinical credibility and the n. Finance needs cash, labor, and what stops if this starts. Front-line managers need the unit list, not a system average that erases them.
How A.11 and A.19 attach to A.5 and A.6
- A.5 tells you whether the SLA, indicator, or system met its own bar.
- A.6 tells you whether people and services are actually working.
- A.11 tells you whether that result is ordinary, better, or worse than a fair comparison.
- A.19 carries the package into a decision.
A green SLA with a worst-quartile effectiveness indicator is an A.11 finding you must not soften in the A.19 room. A worst-quartile satisfaction score with no peer and a 4% response rate is a weak comparison—say so, then recommend a better sample, not a new EHR on that number alone.
Distinctions the exam will punish
- Benchmark versus target versus last month.
- Peer versus celebrity organization.
- Indicator versus vanity count.
- Interpretation versus data dump.
- Recommendation versus the decision itself. You recommend; they decide.
- A.11 employ the comparison versus A.1 measure the organizational goal versus A.5 evaluate against the SLA. Related, not identical.
Scenarios and exam traps
Scenario. A critical-access hospital’s crude inpatient mortality sits above a famous academic medical center’s published rate. Someone wants a “we are worse than National Brand” slide to justify a command-center purchase. A.11 says no: teaching status, transfer-in mix, and hospice coding are not aligned. Compare to like CAHs or to your own risk-adjusted trend. A.19 can still recommend capacity or transfer-process work if those data support it.
Scenario. Barcode scan compliance is 71% against an 85% internal target and 82% among like medical-surgical units in a subscription peer set. Night shift and two units with chronic dead carts dominate the residual. Tickets match the device story (A.6). The A.19 recommendation is restore devices and the swap service, then re-measure—not “buy the next scanning platform because we are below benchmark.”
Scenario. You are on the finance committee calendar for 15 minutes. You bring 36 slides of warehouse extracts and no ask. That fails A.19. Open with the decision (“approve $X to replace 40 carts this quarter, displacing the portal-redesign sprint”), the comparison, the residual, and two options. Leave the extracts in an appendix they did not ask to tour.
Scenario. Portal-login growth is in the top quartile of a vendor’s “digital engagement” chart. The organizational access aim is still 38-day waits. A.11 does not let a vanity engagement benchmark replace the access indicator. A.19 does not tell the board digital is winning.
Scenario. Quality wants you to assert that last year’s interface project caused the length-of-stay drop because both lines moved. Unless you have a design that can support that claim, A.19 language is association plus residual, not causality. Over-claiming is how analysts lose the room and how unsafe programs get funded.
Watch these traps:
- Benchmarking against an unmatched celebrity organization or a vendor marketing average.
- Changing the indicator definition until the comparison is flattering.
- Treating a quartile gap as an automatic purchase.
- Dumping charts with no interpretation, residual, or ask.
- Substituting logins, tickets, or uptime for a care or access indicator in the comparison.
- Inventing a HIMSS-official CPHIMS benchmark table or composite index.
A critical-access hospital’s crude mortality is higher than a famous academic medical center’s published rate. A sponsor wants that comparison used to justify a command-center purchase. What is the best A.11 response?
Scan compliance is 71% versus an 85% target and 82% in a like-unit peer set. Dead carts and night shift explain most of the residual, matching ticket themes. Which A.19 presentation is best?
You have 15 minutes with the finance committee. Which package best satisfies presenting interpretations and recommendations of data analyses?