Reducing CHF Readmissions When the Chart Review Points in Three Directions
A 320-bed community hospital's case management department reviewed 412 CHF discharges from the past year. The 30-day all-cause readmission rate is 26.8% against a national benchmark of 21.6%. Chart review of 60 readmissions found: 68% had no follow-up appointment scheduled within 7 days; 55% were discharged on Friday or Saturday; 41% were on eight or more medications; 33% had a documented note that the patient could not explain the sodium or daily-weight instructions; and only 22% had any weights recorded in the last 10 days before readmission. The department has four RN case managers and two social workers, no dedicated data analyst, and a new electronic health record that can build reports but has no validated readmission dashboard. Leadership wants "the readmission rate down by 5 points in six months" and has asked for a plan. The next PDSA cycle starts in three weeks.
1. Which measures should the department track, and how should process, outcome, and balancing measures be distinguished here?
Read the worked answer
Distinguish process measures (what you do), outcome measures (what happens to patients), and balancing measures (unintended consequences elsewhere). Outcomes: 30-day all-cause readmission rate, 30-day ED visits, and patient-reported confidence in self-management. Process measures tied to the chart-review findings: percentage with a follow-up appointment scheduled before discharge and kept within 7 days; percentage with documented teach-back on daily weights, sodium, and medication changes; percentage with a scale in the home; percentage with pharmacist medication reconciliation; percentage receiving a 48–72 hour follow-up call. Balancing measures: staff overtime, case manager caseload, observation-status shifts (readmissions can fall if patients are held in observation or in the ED), and mortality. With no analyst, define each measure with a numerator, denominator, data source, and owner, and start with two process measures rather than ten. A 5-point drop in six months is ambitious; track process measures monthly and treat the outcome as lagging.
2. Design the first PDSA cycle. What change, what prediction, and what will the team study?
Read the worked answer
Choose the change with the strongest evidence and the least new resources: every CHF patient has a follow-up appointment scheduled before discharge, within 7 days. Run the cycle on one unit for one week. Plan: decide who schedules (the RN case manager), when (by noon on the day before discharge), how (direct scheduling into the clinic through the EHR), and how it is recorded; predict that about half of discharges will have a documented appointment in week one, with failures clustering on Friday and Saturday discharges. Do it. Study: audit every CHF discharge from that week, list the failures, and ask why at the point of failure - no clinic slots, no transportation, patient declined, discharge after clinic hours. Act: if Friday failures dominate, the next cycle adds a Saturday slot or a telehealth visit; if transportation dominates, add a voucher or a mobile clinic option. Keep cycles small, document each, and adapt based on data rather than opinion. A hospital-wide bundle launched at once cannot tell you which component worked.
3. At month three the monthly readmission rate falls to 20.1%. How should the team decide whether this is real improvement?
Read the worked answer
Monthly rates will bounce around even when nothing changes, so do not declare victory on one point. Plot the data on a run chart or p-chart with control limits. With roughly 412 discharges a year, a month holds about 34 discharges; a two-patient difference can swing the percentage by six points, so a single month is dominated by small-number noise. Look for a shift - six or more consecutive points on one side of the median - a trend of five or more consecutive increases or decreases, or points beyond the control limits. Those signal special-cause change; a stable, wiggling line means common-cause variation and requires changing the system, not exhorting staff. Also check for measurement drift: risk adjustment, exclusion criteria, and observation-stay policy. If the drop coincides with a coding change or a new observation practice, it is artifact, not improvement. Report the chart and the context to leadership, not a single number.