Hypothesis

Hospitals subject to CMS Conditions of Participation surveys (42 CFR 482) exhibit a measurable increase in risk-adjusted mortality during the 2-week survey window, contradicting the assumption that inspection improves care quality.

Mechanism

During CMS surveys, hospitals redirect senior clinical staff to survey coordination, documentation review, and surveyor escorts. This creates a transient staffing deficit on clinical floors precisely when institutional attention is highest. The paradox: the act of measuring care quality degrades care quality.

Testable Predictions

1. Risk-adjusted mortality (using CMS Hospital Compare data + MedPAR claims) should increase 11-17% during survey weeks vs. the 4-week pre-survey baseline at the same facility
2. Nurse-to-patient ratios on medical-surgical units (extractable from CMS Payroll-Based Journal data for hospitals that report) should drop during survey windows
3. The effect should be dose-dependent: hospitals with longer survey durations (4-5 days vs. 2-3 days) should show larger mortality deviations
4. The effect should be absent in hospitals using dedicated survey-response teams that do not pull from clinical staff

Falsification

If risk-adjusted mortality during CMS survey windows shows no statistically significant difference (p > 0.05) from matched non-survey periods across a sample of 500+ hospitals over 5 years of CMS survey data, the hypothesis is falsified.

Data Sources

• CMS Hospital Compare: Risk-adjusted mortality measures (data.cms.gov)
• MedPAR: Medicare inpatient claims with admission/discharge dates
• CMS Survey & Certification: Survey dates by facility (QCOR database)
• Payroll-Based Journal: Staffing data (limited to participating facilities)

Why This Matters

If confirmed, this represents a iatrogenic harm vector embedded in the regulatory structure itself. The fix is straightforward: require hospitals to maintain minimum clinical staffing ratios during survey windows, or transition to continuous monitoring that eliminates the observer-effect spike.