Epidemiology & Healthcare Quality Analytics
Age-Standardized Rate & SMR Simpson’s Paradox Formula
Resolve hospital mortality Simpson’s Paradox using Direct Age-Standardized Rate (DASR per 1,000) and Indirect Standardized Mortality Ratio SMR = Observed / Expected.
Catch Simpson’s Paradox when a top referral hospital’s crude mortality looks 65% worse than average!
Comparing unadjusted Crude Mortality Rates SUM(Deaths_i) / SUM(Patients_i) across hospitals or regions routinely triggers Simpson’s Paradox: a world-class tertiary cancer/trauma center that achieves 20% LOWER mortality in EVERY single age bracket still posts a much higher crude death rate simply because 60% of its referred patients are seniors! Epidemiology fixes age confounding with two gold-standard metrics: (1) Direct Age-Standardized Rate DASR_1k = 1,000 * SUMPRODUCT(Std_Pop_Weights, Hospital_Stratum_Rates) / SUM(Std_Pop_Weights) (answering: what rate would this hospital post on the standard national age pyramid?); and (2) Indirect Standardized Mortality Ratio SMR = SUM(Observed_Deaths) / SUMPRODUCT(Hospital_Patients, National_Ref_Rates) (SMR < 1.00 proves fewer deaths than expected for its exact patient case-mix!)!
Verify on 3 age strata Young / Middle / Senior: Standard_Weights B2:B4 = [0.50, 0.30, 0.20], National_Ref_Rates C2:C4 = [0.010, 0.030, 0.100] (National Rate = 34.00 per 1,000), Hospital_Patients D2:D4 = [1,000, 3,000, 6,000] (10,000 total; 60% Senior!), and Observed_Deaths E2:E4 = [8, 72, 480] (560 total; stratum rates F2:F4 = [0.008, 0.024, 0.080] — 20% better in EVERY stratum!). Naive Crude Rate is 56.00 per 1,000 (looks terrible vs 34.00!), whereas Direct Age-Standardized Rate is 27.20 per 1,000 and Indirect SMR = 560 / 700 = 0.8000 (20.00% better than expected!)!
| Hospital Quality Metric (10,000 Patients, 560 Deaths) | Formula | Verified Output |
|---|---|---|
| Confounded Crude Rate (per 1,000) | 1,000 * SUM(E2:E4) / SUM(D2:D4) | 56.00 / 1,000 (Simpson’s Paradox!) |
| Direct Age-Standardized Rate (DASR) | 1,000 * SUMPRODUCT(B2:B4, F2:F4) | 27.20 / 1,000 (vs 34.00 National!) |
| Expected Case-Mix Deaths | SUMPRODUCT(D2:D4, C2:C4) | 700.00 expected deaths |
| Indirect Standardized Mortality Ratio | SUM(E2:E4) / Expected_Deaths | 0.8000 (20% Below Expected!) |
Direct Age-Standardized Rate per 1,000 (DASR) Formula
Normalize by SUM(B2:B4) so Direct Standardization works identically whether standard weights are entered as proportions (1.00) or raw census counts (100,000).
=IF(OR(MIN(B2:B4) < 0, SUM(B2:B4) <= 0, MIN(D2:D4) <= 0), "Invalid Stratum Input", ROUND(1000 * SUMPRODUCT(B2:B4, F2:F4) / SUM(B2:B4), 2))Indirect Standardized Mortality Ratio (SMR = Observed / Expected) Formula
Use SMR whenever small clinic strata have volatile or zero stratum counts: SMR multiplies stable National_Ref_Rates C2:C4 against Hospital_Patients D2:D4!
=LET(obs_d, SUM(E2:E4), exp_d, SUMPRODUCT(D2:D4, C2:C4), IF(OR(MIN(D2:D4) < 0, MIN(C2:C4) < 0, exp_d <= 0), "Invalid Stratum Input", ROUND(obs_d / exp_d, 4)))Try the example
Use the original values or change them to test your understanding.