Table 4

Results of adjusted regression models

Relative differenceAbsolute differenceP valueInteraction test
P value*
Outcomes over the follow-up periodPoint estimate95% CIPoint estimate95% CI
Residential care homes
 Potentially avoidable admissions0.50(0.30 to 0.82)−0.20(−0.28 to –0.07)0.0070.020
 Emergency admissions0.60(0.42 to 0.86)−0.37(−0.54 to –0.13)0.0040.005
 A&E attendances0.57(0.40 to 0.81)−0.57(−0.80 to –0.25)0.0010.008
 Hospital bed days†0.81(0.43 to 1.53)−0.007(−0.022 to 0.021)0.467Count: 0.858
Zero: 0.135
 Outpatient attendances1.28(0.88 to 1.86)0.41(−0.18 to 1.26)0.1690.266
 Deaths outside hospital1.25(0.54 to 2.92)17.3%(−31.8% to 132.7%)0.6080.558
Nursing homes
 Potentially avoidable admissions1.17(0.76 to 1.81)0.03(−0.05 to 0.16)0.484
 Emergency admissions1.15(0.85 to 1.56)0.08(−0.08 to 0.30)0.352
 A&E attendances1.04(0.78 to 1.38)0.03(−0.15 to 0.26)0.806
 Hospital bed days1.07(0.64 to 1.80)0.002(−0.012 to 0.027)0.790
 Outpatient attendances0.97(0.71 to 1.32)−0.06(−0.60 to 0.67)0.840
 Deaths outside hospital1.00(0.40 to 2.49)0.00(−53.1% to 131.9%)0.992
  • All relative differences are rate ratios, apart from deaths outside hospital and deaths, which are ORs. See online supplementary file 2 for a list of covariates each outcome was adjusted for and for model fit statistics. Absolute differences are calculated by first calculating the relative difference, then multiplying the relative difference with the crude rate in the matched control group, and then comparing the resulting rate to the crude rate.

  • *Interaction test of residential versus nursing homes.

  • †The absolute difference in hospital bed days shows the difference in the proportion of hospital bed days out of the period residents were followed up in the study. The interaction model estimating hospital bed days is a Hurdle model and produces separate effect estimates on whether a resident is hospitalised or not (‘Zero’ part) and on the number of hospital bed days following hospitalisation (‘Count’ part).

  • A&E, Accident & Emergency.