10.3 Process Improvement: DMAIC and PDCA

10.3 Process Improvement: DMAIC and PDCA

Quick Answer: Task A.3 is which improvement method fits. DMAIC is Define-Measure-Analyze-Improve-Control for an existing, data-rich process with variation. PDCA/PDSA is a small test of change. Lean removes waste; Six Sigma reduces harmful variation. Do not start at Improve with a purchase. Hold the gain in Control or Act.

Tasks A.1 and A.2 told you how a system is born. Task A.3 is how a living process is improved—emergency-department throughput, admission medication reconciliation, discharge teaching, first-dose turnaround—without pretending a new module is a method. CPHIMS expects you to match the tool to the problem, not to the belt color in the room.

CPHIMS practice questionsPractice questions with detailed explanations

Why method choice is an HIT skill

Informatics is often asked to “fix throughput” by installing a tracker, or to “fix med rec” by turning on a vendor module. Those are solutions looking for a charter. A method forces you to define the problem, measure the current process, and only then change something you can hold.

The exam will offer a vice president with a product and a date. The correct answer is usually the next step in the cycle that was skipped—almost always Measure or Analyze, sometimes Control of a gain that already happened and then drifted.

DMAIC

DMAIC is the Six Sigma backbone for an existing process:

1. Define the problem, the customer, the scope, and the outcome metric. “ED is busy” is not a definition. “Left-without-being-seen is 6.8% and varies by day and attending, harming access” is a definition.
2. Measure the current process with operational definitions. Door-to-doc, door-to-disposition, boarding hours, and left-without-being-seen need timestamps everyone trusts. If the tracker clocks are wrong, you do not have Measure.
3. Analyze causes. Stratify. Look at variation by hour, role, and site. Separate common cause from a special-cause night when the CT scanner died.
4. Improve the few changes that attack those causes. A new bed-management rule, a lab-turnaround path, or a registration change may beat a new screen.
5. Control so the gain stays: standard work, a dashboard with an owner, a response when the metric drifts, and a rule that the old workaround is retired.

DMAIC is heavy. It needs data, time, and a process that already runs every day. Use it for chronic, high-volume problems: left-without-being-seen, medication-order turnaround with wide variation, claim-denial rates by edit, incomplete admission reconciliation by time of day.

Do not use full DMAIC for a two-week checklist test on one hospitalist team. That is PDSA wearing a belt.

PDCA and PDSA

PDCA (Plan-Do-Check-Act) and PDSA (Plan-Do-Study-Act) are the same family. Healthcare quality (IHI Model for Improvement) usually says PDSA. CPHIMS may use either letters. The idea is a small test of change, not a capital project.

1. Plan one change, a prediction, and the data you will look at. “Try a new admission reconciliation checklist with one hospitalist team for two weeks; we predict completeness will rise and time-to-complete will not exceed the current median.”
2. Do the test on a limited scale—one team, one unit, one shift—not the entire enterprise EHR.
3. Check/Study the data and the stories. Did completeness rise? Did night shift invent a bypass? Did pharmacy still retype the list?
4. Act to adopt, adapt, or abandon. Then run the next cycle, or spread with a control plan.

PDSA is the right fit when the change is local, the risk of a small test is acceptable, and you need learning more than a 200-page charter. It is how you test a new triage question, a discharge-teaching sequence, or a quieter CDS prompt before you spend attention hospital-wide.

PDSA is not an EHR replacement method by itself. A multi-year upgrade still needs SDLC (task A.1) and a program. You may run PDSA cycles inside design and implementation—usability tests, training pilots, command-center scripts—but you do not “PDSA” a blood-bank interface into production.

Check is not optional. A cycle that Plans and Does and then “feels better” is not PDCA. It is a go-live party.

Lean waste versus Six Sigma variation

Lean asks what does not add value from the patient’s or the next clinician’s point of view. Classic wastes you will see in HIT and clinical workflow:

• Transportation / motion — walking to a distant printer for discharge instructions; pharmacy trips from a satellite to a central printer to the tube station.
• Waiting — boarding, pending lab status that nobody can see, an after-visit summary locked behind an unclosed note.
• Overprocessing / extra clicks — fourteen required fields before a sepsis set will open; re-entering home medications that already exist.
• Inventory — batching discharge prescriptions until 16:00; piles of printed MARs.
• Defects — wrong home med, duplicate imaging, a tracker that shows the wrong bed.
• Unused skill — pharmacists retyping lists instead of reconciling clinical issues.

If everyone takes the same bad path and variation is small, you have a waste problem. Remove steps, bring the printer closer, show the status, stop the re-entry. A Six Sigma theater of control charts will delay an obvious motion waste.

Six Sigma asks why the outcome varies and which variation is harmful. Left-without-being-seen that is 2% on Tuesday mornings and 11% on Monday nights is a variation problem. First-dose antibiotic time that is protocol-true at 10:00 and chaotic at 02:00 is a variation problem. DMAIC and statistical thinking belong here.

Many HIT problems are mixed. Map waste first (Lean, often with the visuals in 10.4). If the remaining pain is spread and data-rich, run DMAIC. Do not start a black-belt project to discover that everyone walks to the same broken printer.

When each fits HIT

Rules the exam will punish you for breaking

• Do not start at Improve or Do with a purchase.
• Do not use DMAIC as bureaucracy for a two-week local test.
• Do not use PDSA as a substitute for SDLC on a safety-critical enterprise change.
• Do not confuse Lean (waste) with Six Sigma (variation).
• Do not skip Control. Standard work, an owner, and a metric are part of the method.
• Do not measure vanity activity (meetings held, modules installed) instead of the outcome you Defined.

How to read a CPHIMS improvement stem

1. Is this a new system (SDLC) or an existing process (A.3)?
2. Is the pain waste, variation, or both?
3. How big is the proposed change, and how much trusted data exist?
4. What step was skipped—usually Measure, Analyze, or Control?
5. Prefer the next honest step in the matching cycle over the product.

Scenarios and exam traps

Scenario — ED throughput. Left-without-being-seen varies widely by day and attending. The hospital has months of timestamps. A VP wants to buy a new tracker next week. Use DMAIC. Define LWBS, measure the clocks, analyze boarding and lab delay, then improve. The tracker is a candidate intervention, not the method. A PDSA on the purchase meeting is not analysis.

Scenario — med-rec checklist. Pharmacy and hospitalists want to test a new admission checklist with one team for two weeks. Use PDSA: plan the prediction, do the test, study completeness and time, act. A six-month DMAIC charter before anyone may try the card is method theater.

Scenario — discharge teaching waste. Everyone walks to the same distant printer and repeats four EHR clicks. Variation is small. Use Lean. Move printing, cut clicks, stop batching. Do not open a Six Sigma project to prove walking is slow.

Watch these traps:

1. Starting at Improve with a module or tracker.
2. Using DMAIC for a tiny test, or PDSA for an enterprise safety-critical replacement.
3. Calling all delay “variation” when the waste is shared motion or extra clicks.
4. Planning and Doing without Check/Study.
5. Forgetting Control so the gain disappears when the team leaves.
6. Counting installations instead of LWBS, completeness, or time-to-dose.

Task A.3 is match-the-method. Section 10.4 is how you see the process those methods are trying to change.