Table 1

The five multimorbidity indices

Multimorbidity index weights*
Index 1—absolute countIndex 2—individual unweighted conditionsIndex 3—weighted LASSOIndex 4—weighted AICIndex 5—weighted BMA
Mood disorder191.22.81.8
Anxiety17.82.4
Osteoarthritis115.35.54.7
Rheumatoid arthritis132.8
Other type of arthritis13.21.1
Asthma12.5
COPD113.83.84.54.1
Hypertension110.63.63.3
Heart disease114.67.44.6
Angina14.711.54
Myocardial infarction19.62.23.03.9
Peripheral vascular disease110.32.13.3
Stroke or TIA144
Gastrointestinal ulcer111.92.23.72.4
Bowel disorder15.81.91.9
Kidney disease or failure113.61.74.24.5
Hypothyroidism14.11.31
Hyperthyroidism16.72.2
Diabetes17.81.62.4
Osteoporosis11.42
Cataracts12.7
Glaucoma1−3.2
Macular degeneration11.5
Epilepsy119.15.8
Parkinson’s disease133.210.27.9
Multiple sclerosis123.37.1
Migraine11
Allergies12.71
Skin cancer: melanoma1−6.7
Skin cancer: non-melanoma1−2.9
Solid cancer16.422.9
Haematological and soft cancer130.69.5
Ill-defined cancer1−5.1
Current smoker14.41.3
Regular drinker1−10.5−1.9−3.3
Internal predictive validity of the regression models (in the validation dataset)†‡
Nagelkerke’s R2§0.060.070.060.070.06
C-statistic (95% CI)§0.67 (0.65 to 0.68)0.68 (0.67 to 0.70)0.67 (0.65 to 0.68)0.68 (0.66 to 0.69)0.66 (0.65 to 0.68)
Pearson correlation (95% CI)§¶0.88 (0.84 to 0.92)0.89 (0.85 to 0.92)0.77 (0.69 to 0.83)0.89 (0.85 to 0.92)0.86 (0.81 to 0.90)
  • *Index 1 is based on a regression model including the absolute count of chronic conditions as a single independent variable. Index 2 is based on a regression model where each chronic condition was treated as a dichotomous independent variable. Indices 3–5 are based on regression models that began with the same independent variables as those used for index 2, but then involved variable selection using LASSO, stepwise selection based on the Akaike information criterion (AIC), or Bayesian model averaging (BMA), respectively.

  • †There are many estimates for each measure due to multiple imputation. We report the median estimate.

  • ‡For the model reference model including only age and sex, R2=0.03, C-statistic=0.62 (95% CI: 0.60, 0.64), and correlation coefficient=0.83 (95% CI: 0.74, 0.89).

  • §These are based on the model including the index, sex, age and an interaction between age and the index.

  • ¶Correlation between the predicted and the observed probability of the outcome, using 130 quantiles.

  • COPD, Chronic obstructive pulmonary disease; LASSO, least absolute shrinkage and selection operator; TIA, transient ischaemic attack.