Quality Improvement Methods and Process Redesign

Introduction

Healthcare quality improvement (QI) is the disciplined, continuous effort to make care safer, more effective, timely, efficient, equitable and patient-centred. Hospitals are complex systems in which outcomes depend not only on clinical knowledge, but also on communication, staffing, equipment, information systems, procurement, physical layout and organisational culture. Consequently, poor performance is rarely explained by one careless individual. It is more often produced by processes that allow errors, delays, duplication, variation or waste.

This chapter presents practical methods for redesigning healthcare processes. It focuses on PDCA/PDSA, Lean, Six Sigma and DMAIC, Kaizen, 5S, process mapping, clinical audit, benchmarking and root-cause analysis. These methods are complementary rather than mutually exclusive. A hospital may use process mapping to understand a medication pathway, root-cause analysis to investigate a serious incident, Lean to remove unnecessary movement, and PDSA cycles to test a revised process.

Quality improvement differs from inspection. Inspection identifies whether a standard has been met; improvement changes the system so that the standard is met more reliably. It also differs from research. Research seeks generalisable new knowledge, whereas QI usually tests whether an established intervention works in a particular service. Nevertheless, rigorous measurement, ethical conduct and transparent documentation are essential in both.

Learning objectives

By the end of this chapter, the learner should be able to:

1. Explain the purpose and distinctive features of major QI methods.
2. Select appropriate process-redesign and analytical tools for healthcare problems.
3. Construct a process map and identify waste, bottlenecks and failure points.
4. Apply PDSA, DMAIC, clinical audit, benchmarking and root-cause techniques.
5. Use basic quality measures to assess variation and improvement.
6. Design sustainable interventions that protect patients, staff and information.
7. Critically evaluate the limitations, ethical implications and governance requirements of QI.

1. Foundations of healthcare improvement

Quality as a system property

A useful starting point is Donabedian’s structure–process–outcome model. Structure includes buildings, staff, equipment, financing and information systems. Process concerns what is done for and with patients, such as assessment, prescribing, handover and discharge. Outcomes include mortality, complications, functional status, patient experience and cost. A process measure may improve before an outcome changes; this is not necessarily failure, because outcomes are influenced by many variables and may require longer observation.

The Institute of Medicine traditionally describes quality through dimensions such as safety, effectiveness, patient-centredness, timeliness, efficiency and equity. These dimensions can conflict. For example, reducing length of stay may improve timeliness and efficiency but harm safety if discharge planning is inadequate. Improvement therefore requires balanced measures rather than a single target.

Variation and waste

Variation is the difference in performance between patients, clinicians, shifts, locations or time periods. Some variation is appropriate because patients differ clinically. Unwarranted variation occurs when outcomes or treatment differ without a defensible clinical reason. Standardisation should reduce unwarranted variation while preserving professional judgement for legitimate exceptions.

Waste is activity that consumes resources without adding value for the patient. Lean healthcare commonly identifies eight forms:

A process can be clinically necessary yet poorly designed. Improvement asks two questions: “Does this step contribute to the patient’s desired outcome?” and “Can it be performed more reliably, safely and efficiently?”

Selecting measures

A balanced improvement project normally uses:

• Outcome measures, such as falls per 1,000 occupied bed-days.
• Process measures, such as the percentage of eligible patients receiving falls-risk assessment within four hours.
• Balancing measures, such as staff workload, unintended readmissions or adverse effects caused by the intervention.

For a count-based rate, the general expression is:

The multiplier may be 100, 1,000 or 100,000, depending on event frequency. The numerator must be included in the denominator’s population and the time period must be specified. For example, 12 medication errors among 4,000 medication administrations equals 3 errors per 1,000 administrations. A rate should not be compared across units unless definitions, ascertainment, exposure and case mix are sufficiently comparable.

For a proportion:

A proportion has no unit, although it is reported as a percentage. Measures must have an operational definition: precisely who is included, what counts as an event, the data source, the collection frequency and the responsible owner.

2. PDCA and PDSA: learning through controlled change

The Plan–Do–Check–Act (PDCA) cycle and the Plan–Do–Study–Act (PDSA) cycle are iterative approaches to improvement. Both move from understanding a problem to testing a change and deciding what to do next. PDSA is particularly useful when a team wants to study results in detail rather than merely check compliance.

The four stages

Plan: Define the problem, aim, population, measures, predicted result and proposed change. Identify risks and decide how data will be collected.

Do: Test the change on a small scale, documenting what actually happened, deviations, workload and unintended effects.

Study or Check: Compare results with the prediction. Examine quantitative data and qualitative observations. Ask whether the change, rather than another factor, plausibly produced the result.

Act: Adopt the change if it is effective and safe, adapt it if it requires modification, or abandon it if it fails. Begin another cycle with the learning incorporated.

A strong aim statement is specific: “Within twelve weeks, increase the percentage of adult medical inpatients receiving medication reconciliation within 24 hours from 62% to 90%, without increasing discharge delays.” It identifies the population, measure, baseline, target and time frame.

Strengths and limitations

PDSA encourages rapid learning and staff participation. Small tests reduce risk and reveal practical barriers before full implementation. However, teams sometimes label ordinary implementation as PDSA without making predictions, collecting data or studying results. A before-and-after change may also be influenced by seasonality, staffing or concurrent initiatives. Run charts, repeated measurements and transparent documentation improve credibility.

3. Lean and Kaizen in healthcare

Lean originated in industrial production but its principles can be adapted to services. Lean does not mean simply reducing staff or costs. Its central purpose is to increase value from the patient’s perspective by removing activities that do not contribute to safe, effective care.

Lean principles

1. Define value: Determine what patients and users consider important.
2. Map the value stream: Display all steps, information flows, delays and handoffs.
3. Create flow: Reduce interruptions, queues, batching and unnecessary movement.
4. Establish pull: Provide work or supplies when needed rather than producing excessive inventory.
5. Pursue perfection: Continue learning and eliminate root causes of problems.

Lean tools include visual management, standard work, workload balancing, just-in-time supply, error-proofing and the five-whys technique. Lean must be applied carefully in clinical environments: a patient is not a factory product, demand is variable, and safety sometimes requires deliberate redundancy.

Kaizen means continuous, incremental improvement. It emphasises frequent small changes led by the people closest to the work. A ward-based Kaizen event might reorganise supplies, clarify escalation criteria and redesign a handover board over several days. Kaizen is powerful for engagement but cannot substitute for major capital investment, workforce planning or clinical redesign when the underlying problem is structural.

5S workplace organisation

5S is a method for organising the workplace:

• Sort: Remove unnecessary items.
• Set in order: Locate essential items so they are easy to find and return.
• Shine: Clean and inspect the area.
• Standardise: Establish agreed layouts, labels and routines.
• Sustain: Audit, coach and improve the arrangement.

In an emergency trolley, 5S may reduce search time and omissions. It should not be treated as cosmetic housekeeping. Labels, standard locations and stock checks must be designed with infection prevention, accessibility, medication security and human factors in mind.

4. Six Sigma and DMAIC

Six Sigma is a data-oriented approach to reducing defects and unwanted variation. In healthcare, the term “defect” must be defined carefully—for example, a specimen rejected because of incorrect labelling or a referral lacking required information. A clinical outcome is not automatically a defect merely because it is undesirable; definitions must account for patient risk and clinical appropriateness.

DMAIC is the principal improvement sequence for an existing process:

Define

Develop a problem statement, scope, customer or patient requirements, project charter, stakeholders and improvement aim. Avoid an overly broad problem such as “improve the emergency department.” A bounded problem might be “reduce time from triage completion to first clinician assessment for category-three patients.”

Measure

Describe the current process and establish a reliable baseline. Validate data definitions, sampling, completeness and measurement consistency. A poor measurement system can make a stable process appear variable.

Analyse

Identify patterns, bottlenecks and likely causes. Tools include Pareto charts, stratification by shift or location, cause-and-effect diagrams, scatter plots and statistical process-control charts. Correlation is not proof of causation; clinical and operational knowledge remains essential.

Improve

Generate and test solutions. Prefer changes that redesign the system, such as standardised electronic prompts, simplified forms, visual cues or appropriate automation, rather than relying only on memory and training.

Control

Maintain gains through standard work, ownership, monitoring, escalation rules, staff induction and periodic review. A control plan identifies the measure, frequency, threshold, responsible person and response to deterioration.

A Pareto display ranks causes by frequency or impact, often revealing that a small number of categories account for a large proportion of problems. The “80/20 rule” is a heuristic, not a law; data should determine whether concentration exists.

Basic process capability concepts

When a process has a measurable continuous output, its mean and spread can be examined. The sample mean is:

where xᵢ is observation i and n is the number of observations. The sample standard deviation is:

The denominator n − 1 is used when estimating population variability from a sample, assuming observations are reasonably independent and the measure is appropriate for this use. These statistics describe data; they do not by themselves establish clinical significance or causation.

Control limits on a statistical process-control chart are commonly estimated as:

Here σ represents the process standard deviation under the selected charting method. The limits describe expected process variation, not regulatory or clinical specification limits. A point outside a limit or a non-random pattern may indicate special-cause variation, but chart rules and assumptions must be defined in advance. Proportions, counts and measurements require different chart types; an individuals chart should not automatically be used for every dataset.

5. Process mapping and redesign

Process mapping makes work visible. A basic map lists activities in sequence, decision points, handoffs, information inputs, outputs and delays. A cross-functional “swim-lane” map places each activity in the lane of the responsible role or department, exposing communication gaps and duplicated work.

A practical mapping procedure is:

1. Select a clearly bounded process and define its start and end.
2. Assemble people who perform the work, including patients where possible.
3. Observe the process directly rather than relying only on policy documents.
4. Record actual steps, queues, rework, interruptions and workarounds.
5. Mark safety-critical points, failure modes and information handoffs.
6. Measure time, distance, defects and volume where useful.
7. Design a future-state map and test it with stakeholders.

Distinguish cycle time, the time required to perform work, from lead time, the total elapsed time including waiting. A patient may require only 30 minutes of active assessment within a four-hour pathway. Reducing lead time may involve queue management rather than making clinical assessment faster.

Process redesign should follow the hierarchy of controls where applicable. First eliminate unnecessary steps; then simplify and standardise; then use engineering or technological controls; finally rely on training, reminders and individual vigilance. Training alone is the weakest solution when the process is inherently confusing.

Failure Modes and Effects Analysis

Failure Modes and Effects Analysis (FMEA) is prospective. The team lists possible failure modes, their effects and causes, then prioritises risks before harm occurs. Traditional scoring may use severity, occurrence and detectability ratings, but multiplying subjective scores can create false precision. Teams should record rating definitions, discuss high-severity failures separately and consider patient harm even when frequency is low.

6. Clinical audit and benchmarking

Clinical audit is a structured cycle in which current practice is measured against explicit criteria, gaps are identified, changes are implemented and practice is re-measured. Its sequence is:

1. Select a high-priority topic.
2. Define evidence-based criteria and an acceptable standard.
3. Identify the population and sample.
4. Collect valid, confidential data.
5. Compare performance with the standard.
6. Implement an action plan.
7. Re-audit after a suitable interval.

An audit criterion must be observable and measurable. “Good pain management” is weak; “eligible postoperative patients have pain reassessed within 60 minutes after analgesia” is more suitable, provided the criterion is clinically justified.

Benchmarking compares performance with a reference point: another unit, a national dataset, a historical baseline or a recognised standard. Internal benchmarking supports learning between comparable departments; external benchmarking can reveal relative performance but may be misleading when definitions, case mix, resources or coding practices differ. Benchmarking should prompt questions, not encourage blame or superficial ranking.

Risk adjustment may be necessary when comparing outcomes among different patient populations. At an advanced-diploma level, the key principle is that crude rates should not be interpreted as evidence of superior care without considering patient complexity, denominator integrity and data quality.

7. Root-cause analysis and learning from harm

Root-cause analysis (RCA) is a retrospective investigation of how an incident occurred and why system defences failed. It should not be a search for one “root” cause or a disguised disciplinary hearing. Most incidents involve interacting conditions: workload, unclear policy, equipment design, communication, supervision and latent organisational decisions.

A robust RCA includes:

• a factual timeline;
• statements from patients, families and staff where appropriate;
• identification of actual and potential harm;
• analysis of task, team, technology and organisational factors;
• examination of barriers that should have prevented the event;
• recommendations linked to owners and deadlines;
• follow-up to determine whether risk has actually fallen.

The five-whys method asks why repeatedly until the team reaches actionable system conditions. It is useful for simple causal chains but can oversimplify complex events. Fishbone diagrams organise possible causes under headings such as people, methods, equipment, environment, measurement and management. A fault-tree approach begins with an undesired event and works backward through combinations of contributing failures.

Recommendations should be specific and strong. Examples include forcing functions, standardised equipment connections, independent verification for high-risk medicines, redesigned electronic fields and improved staffing escalation. “Remind staff to be careful” is a weak recommendation because it does not reliably change system conditions.

8. Integrated application: illustrative case

Background

The medical ward of an illustrative 250-bed hospital reported delayed discharge and frequent patient complaints. A review suggested that discharge summaries were often incomplete and prescriptions were not ready when patients were medically fit.

Problem and baseline

The improvement team defined the process from the clinician’s decision that a patient was medically fit to the patient leaving the ward. Over two weeks, 80 discharges were observed. The median lead time was 5.2 hours. Only 48 of 80 patients had discharge medication available within two hours of the decision. Twelve patients experienced a delay exceeding eight hours. The team also noted repeated telephone calls to pharmacy and duplicated data entry.

Analysis

A swim-lane map showed four major handoffs: medical team to nursing staff, medical team to pharmacy, pharmacy to transport and ward to family or receiving service. A Pareto review of 30 documented delays found: incomplete medication information (12), late senior decision (8), pharmacy queue (6), transport coordination (3) and family availability (1). An RCA of two delayed discharges identified an electronic form that did not clearly distinguish “draft” from “final,” and no standard time for notifying pharmacy.

Intervention

Using PDSA, the team tested a morning “estimated discharge” huddle for five patients. It introduced a standard electronic discharge checklist, a clear finalisation field, early pharmacy notification and a visual board showing outstanding actions. The pilot was reviewed daily. Staff found that the first version created duplicate documentation, so the checklist was simplified in the second cycle. Pharmacy agreed to prioritise prescriptions marked as clinically urgent, subject to safe verification.

Outcome and control

After six weeks, 92 of 110 discharges had medication available within two hours, compared with 48 of 80 at baseline. The proportion increased from 60% to 83.6%:

The result was encouraging but not conclusive evidence of causation. The team monitored balancing measures: medication errors, overtime, readmissions within seven days and staff-reported workload. No increase in medication errors was observed during the pilot, although longer monitoring was required. The control plan assigned weekly review to the ward manager and monthly audit to the quality department. New staff received orientation, and the electronic form was locked against ambiguous status labels.

Learning points

The case illustrates the value of combining methods. Mapping revealed handoffs and duplication; Pareto analysis prioritised causes; RCA examined system weaknesses; PDSA enabled a low-risk test; and control planning supported sustainability. It also shows why a percentage improvement should be interpreted alongside safety, workload and patient outcomes.

9. Implementation, sustainability and governance

Improvement fails when the technical solution is not adopted in daily work. Implementation requires visible leadership, a multidisciplinary team, patient and family involvement where feasible, protected time, transparent communication and clear accountability. The team should identify those affected by the change, including groups who may experience unequal access or burden.

Sustainability means that improved performance continues after the project team withdraws. Important mechanisms include standard operating procedures, competency-based induction, reliable supplies, compatible information systems, audit feedback, ownership by line managers and escalation when performance deteriorates. A run chart or control chart is more informative than a single favourable measurement.

QI data may contain personal or sensitive health information. Governance requirements include minimum necessary access, secure storage, lawful data use, de-identification where possible and appropriate approval. Patients should not be exposed to unreasonable risk merely because a project is labelled improvement. Where an activity involves research questions, experimental interventions or generalisable knowledge, institutional research review may be required.

Equity must be assessed explicitly. A pathway that improves the average waiting time may worsen access for patients with limited language proficiency, disability, poverty or unstable housing. Data should be stratified where ethically and statistically appropriate, while protecting privacy. Automation and artificial intelligence can support risk prediction, scheduling, documentation and anomaly detection, but they can also reproduce biased data, obscure reasoning and create alert fatigue. Human oversight, validation, explainability appropriate to the use, cybersecurity and monitoring for differential performance are essential.

Challenges and limitations

No improvement method is universally superior. Lean may underemphasise clinical complexity if applied mechanically. Six Sigma can become data-heavy and slow when a simple local change is sufficient. PDSA can be weakened by poor measurement or uncontrolled concurrent changes. Clinical audit may encourage compliance with outdated standards. Benchmarking can generate unfair comparisons. RCA may produce shallow recommendations if the team focuses on individual behaviour. 5S may deteriorate when workload, space or supply systems are not addressed.

Common failure modes include selecting a solution before understanding the problem, using unreliable data, setting targets that encourage gaming, ignoring balancing measures, overloading staff with simultaneous projects and declaring success too early. Effective improvement is proportionate: use the least complex method that can answer the question, but apply enough rigour to protect patients and support credible learning.

Best-practice checklist

Before launching a project, confirm that the team has:

• a patient-important problem and defined scope;
• baseline data with clear denominators;
• an aim, theory of change and balanced measures;
• representation from all relevant roles and patient perspectives;
• a process map based on observation;
• assessment of risks, equity and information governance;
• small, documented tests before full implementation;
• named owners, review dates and escalation rules;
• a plan for sustaining and spreading the improvement.

Chapter Summary

Quality improvement redesigns healthcare systems so that safe and effective care is delivered more reliably. PDCA/PDSA supports iterative learning through prediction, testing, study and adaptation. Lean focuses on patient value, flow and removal of waste; Kaizen promotes continuous staff-led improvement; and 5S organises workplaces for reliability. Six Sigma and DMAIC provide a structured, data-oriented approach to reducing defects and variation. Process maps reveal steps, queues and handoffs, while FMEA anticipates failures. Clinical audit compares practice with explicit standards and closes the audit cycle through re-measurement. Benchmarking supports learning but requires comparable definitions and case mix. RCA examines how system defences failed after an incident and should produce strong, actionable recommendations. Sustainable improvement depends on valid measures, multidisciplinary participation, governance, equity, human factors, leadership and ongoing control. No method replaces judgement: tools must be selected according to the problem, evidence and risk.

Key Terms

Quality improvement: Systematic activity intended to improve healthcare processes and outcomes.

PDCA: Plan–Do–Check–Act cycle for iterative improvement.

PDSA: Plan–Do–Study–Act cycle emphasising learning from tests.

Lean: Approach focused on patient value, flow and waste reduction.

Waste: Activity that consumes resources without adding value for the patient.

Kaizen: Continuous, incremental improvement involving people who perform the work.

5S: Sort, Set in order, Shine, Standardise and Sustain.

Six Sigma: Data-oriented approach to reducing defects and process variation.

DMAIC: Define, Measure, Analyse, Improve and Control.

Process map: Visual representation of activities, decisions, handoffs and delays.

Cycle time: Time required to perform a process activity.

Lead time: Total elapsed time, including waiting.

Clinical audit: Measurement of practice against explicit criteria followed by action and re-audit.

Benchmarking: Comparison with a reference standard or comparable organisation.

Root-cause analysis: Structured retrospective examination of contributing system factors.

FMEA: Prospective analysis of possible failure modes and their effects.

Outcome measure: Indicator of the result experienced by patients or the system.

Process measure: Indicator of whether a desired activity is performed.

Balancing measure: Measure used to detect unintended consequences.

Special-cause variation: Non-routine variation attributable to a specific influence.

Review Questions

Short Answer Questions

1. Distinguish an outcome measure from a process measure and give one healthcare example of each.
2. What are the four stages of the PDSA cycle?
3. List four forms of waste identified in Lean healthcare.
4. What is the purpose of a swim-lane process map?
5. Why should recommendations from root-cause analysis address system conditions rather than blame individuals?

Long Answer Questions

1. Explain how Lean, Six Sigma/DMAIC and PDSA differ, and describe how they could be combined in one hospital improvement project.
2. Discuss the role of clinical audit in improving compliance with evidence-based care, including the importance of re-audit.
3. Explain the principles of 5S and evaluate its possible benefits and limitations in a hospital ward.
4. Describe how a team should design outcome, process and balancing measures for a medication-safety project.
5. Discuss the governance, ethical and equity considerations involved in using patient data for quality improvement.

Analytical / Application Questions

1. A ward’s specimen-rejection rate rises after a new electronic ordering system is introduced. Design a DMAIC-based investigation, specifying likely data sources, measures and possible interventions.
2. A hospital reduces average emergency-department waiting time, but complaints from patients with limited English increase. Analyse possible causes and propose an improvement plan that includes equity and balancing measures.
3. Following a wrong-dose medication incident, develop an RCA approach that distinguishes active errors from latent system conditions and produces at least three strong recommendations.

Multiple Choice Questions with Answers

Question 1

Which measure best represents a process measure for medication reconciliation?

A. Percentage of eligible patients reconciled within 24 hours

B. Number of medication-related deaths

C. Patient satisfaction with discharge

D. Average length of stay

Correct Answer: A. Percentage of eligible patients reconciled within 24 hours

Question 2

In the PDSA cycle, what is the main purpose of the Study stage?

A. To permanently implement the intervention

B. To compare results with predictions and interpret learning

C. To assign disciplinary responsibility

D. To purchase new equipment

Correct Answer: B. To compare results with predictions and interpret learning

Question 3

Which Lean waste is illustrated when nurses repeatedly walk to another ward to obtain routine supplies?

A. Overproduction

B. Inventory

C. Motion

D. Defect

Correct Answer: C. Motion

Question 4

What is the correct sequence of DMAIC?

A. Define, Measure, Analyse, Improve, Control

B. Develop, Monitor, Audit, Implement, Check

C. Detect, Manage, Assess, Inspect, Communicate

D. Design, Act, Measure, Improve, Compare

Correct Answer: A. Define, Measure, Analyse, Improve, Control

Question 5

Which activity is the primary purpose of clinical audit?

A. Comparing practice with explicit criteria and implementing corrective action

B. Replacing all professional judgement with a fixed protocol

C. Ranking staff according to individual performance

D. Collecting data without changing practice

Correct Answer: A. Comparing practice with explicit criteria and implementing corrective action

Question 6

A process has 12 errors in 4,000 medication administrations. What is the error rate per 1,000 administrations?

A. 0.3

B. 3

C. 30

D. 300

Correct Answer: B. 3

Question 7

Which is generally the strongest improvement intervention?

A. Asking staff to remember a warning

B. Sending a reminder email

C. Redesigning the system to prevent or force detection of the error

D. Providing an optional poster

Correct Answer: C. Redesigning the system to prevent or force detection of the error

Question 8

What is a key limitation of benchmarking?

A. It cannot use numerical data

B. It always prevents learning between organisations

C. Comparisons may be misleading when definitions or case mix differ

D. It is only applicable to manufacturing

Correct Answer: C. Comparisons may be misleading when definitions or case mix differ

Question 9

Which statement best describes root-cause analysis?

A. A disciplinary process focused on identifying the careless person

B. A retrospective examination of contributing factors and failed defences

C. A method used only to calculate financial costs

D. A prospective method for organising stock

Correct Answer: B. A retrospective examination of contributing factors and failed defences

Question 10

What is the main function of a balancing measure?

A. To replace all outcome measures

B. To detect unintended consequences of an intervention

C. To determine staff salaries

D. To prove that correlation equals causation

Correct Answer: B. To detect unintended consequences of an intervention