Evidence-based design recommendations for prevalence studies on multimorbidity: improving comparability of estimates

Table 4 Variables and effect estimates from the model

Characteristics	Categories	Effect estimate	95% CI
Mean age	Years	0.052	0.044, 0.061
Number of age groups	1	0	–
	2	−2.7	−3.69, −1.71
	3	0.391	−0.14, 0.92
	4	0.474	−0.10, 1.05
	5	0.102	−0.47, 0.67
	6	0.001	−0.87, 0.87
Disease classification	Names of specific disease/disease groups	0	–
	Diseases based on ICD-10 or ICD-9 codes	1.26	0.039, 2.49
	Diseases based on ICPC-2 or CIRS	−0.789	−1.64, 0.067
No. of diseases in the classification	5–9	0	-
	10–24	0.516	0.0017, 1.03
	25–74	1.22	0.43, 2.01
	≥75	0.806	−0.47, 2.08
Setting	General population	0	–
	Primary care practice	0.015	−0.70, 0.73
	Hospital/nursing home	1.41	0.39, 2.44
	Health insurance	1.05	−0.66, 2.76
Data source	Self-report	0	–
	Medical record	−0.863	−1.71, −0.019
	Self-report + medical record	−0.75	−1.43, −0.071
	Administrative data	−0.497	−2.39, 1.39
Data collection period	Up to one year	0	–
Data collection period	One year or more	−0.349	−0.79, 0.09
Data reporting quality	P2+ given in paper	0	–
	P2+ calculated from paper	−0.4	−0.94, 0.14
	P2+ read from graph in paper	−1.76	−2.52, −0.99
Constant		−3.305	−4.01, −2.60

Legend: Random effects model with response: logit P2+, sampling weights inverse to binomial variance of logit P2+; average sampling weight: 0.0182. N = 108; adjusted R2 = 70.6%; residual variance τ² = 0.5812; overall F(20,87) = 12.61; p < 0.00005
Categories with “Effect estimate” = 0 are reference categories. Gender analyses were done using a different data set (see text). Mean age: Effect estimate gives the change in logit P2+ when changing mean age by one year. Other variables: Effect estimate quantifies the effect on logit P2+ of going from the reference category to the category of interest – e.g., going from one age group to two changes logit P2+ by −2.70, when keeping all other variables fixed (also see Discussion)

ISSN: 1478-7954