"The Lidkoping Accident Prevention Programme -- a community approach to preventing childhood injuries in Sweden" by Svanstrom L, Ekman R, Schelp L, and Lindstrom A. Injury Prevention 1995 1: 169-172.

Abstract

Objectives -- In Sweden about 100 children 0-14 years die from accidental injuries every year, roughly 40 girls and 60 boys. To reduce this burden the Safe Community concept was developed in Falkoping, Sweden in 1975. Several years later a second programme was initiated in Lidkoping. The objectives of this paper are to describe the programme in Lidkoping and to relate it to changes in injury occurrence.

Setting -- The Lidkoping Accident Prevention Programme (LAPP) was compared with four bordering municipalities and to the whole of Skaraborg County.

Methods -- The programme included five elements: surveillance, provision of in formation, training, supervision, and environmental improvements. Process evaluation was based mainly on notes and reports made by the health planners, combined with newspaper clippings and interviews with key people. Outcome evaluation was based on information from the hospital discharge registry.

Results -- In Lidkoping there was an on average annual decrease in injuries leading to hospital admissions from 1983 to l991 of 2.4% for boys and 2.1% for girls compared with a smaller decline in one comparison area and an increase in the other.

Conclusions -- Because the yearly injury numbers are small there is a great variation from year to year. However, comparisons over the nine year study period with the four border communities and the whole of Skaraborg county strengthen the impression that the programme had a positive effect. The findings support the proposition that the decrease in the incidence of childhood injuries after 1984 could be attributed to the intervention of the LAPP. Nevertheless, several difficulties in drawing firm conclusions from community based studies are acknowledged and discussed.

Table 2: Incidence of hospitalized injuries (rates/1000 under 14 years) in Lidkoping (intervention area) and comparison areas by year and gender (Pop. =area population)

	INTERVENTION AREA						COMPARISON AREA 4 BORDER MUNICIPALITIES
	Girls			Boys			Girls			Boys
Year	#	Pop.	Rate	#	Pop.	Rate	#	Pop.	Rate	#	Pop.	Rate
1983	34	3247	10.5	59	3356	17.6	34	4118	8.3	52	4318	12.0
1984	48	3200	15.0	44	3271	13.5	28	4046	6.9	58	4203	13.8
1985	32	3140	10.2	41	3265	12.6	34	4059	8.4	47	4175	11.3
1986	26	3092	8.4	55	3252	16.9	27	4018	6.7	57	4146	13.7
1987	38	3056	12.4	39	3208	12.2	38	4018	9.5	66	4162	15.9
1988	22	3016	7.3	43	3204	13.4	17	4049	4.2	52	4163	12.5
1989	40	3006	13.3	62	3232	19.2	44	4113	10.7	59	4163	14.1
1990	28	3072	9.1	34	3287	10.3	46	4179	11.0	59	4304	13.7
1991	30	3160	9.5	37	3378	11.0	30	4235	7.1	53	4355	12.2
Mean			10.6			14.1			8.1			13.2
95% CI			8.7 to 12.5			11.5 to 16.7			6.4 to 9.8			12.1 to 14.3
beta			-0.3			-0.4			0.2			0.1
%change per year			-2.1			-2.4			2.2			0.6

(Since Skaraborg county INCLUDES the intervention area, we will not analyze the following data in the course)

	COMPARISON AREA: SKARABORG COUNTY
	Girls			Boys
Year	#	Pop.	Rate	#	Pop.	Rate
1983	228	26202	8.7	349	27464	12.7
1984	227	25818	8.8	356	26998	13.2
1985	214	25519	8.4	347	26690	13.0
1986	220	25296	8.7	363	26470	13.7
1987	239	25151	9.5	374	26325	14.2
1988	196	25079	7.8	327	26391	12.4
1989	221	25153	8.8	330	26602	12.4
1990	237	25506	9.3	346	27047	12.8
1991	205	25977	7.9	314	27552	11.4
Mean			8.7			12.9
95% CI			8.3 to 9.1			12.3 to 13.5
beta			-0.1			-0.2
%change per year			-0.3			-1.0

======================

Layout of input in SAS program file ... see web page

(if clicking on the link doesn't work, then right click; save to local disk; and bring,
or cut and paste, program and data into the SAS editor yourself)

Data on rates and population sizes are taken directly from Table 2 of article;

# of injuries reconstructed by JH as nearest integer to population x rate

(1) year
(2) rate
(3) population
(4) Number of injuries
(5) gender*
(6) area**

(1)  (2)   (3)  (4) (5) (6)

1983 10.5  3247  34  0   1

1983 17.6  3356  59  1   1

etc.

1991 11.0  3378  37  1   1

1983  8.3  4118  34  0   0

..

1991 12.2  4355  53  1   0

1991 11.4 27552 314  1   2

* gender: 0 = females; 1 = males

** area 0 = 4 border municipalities; 1 = intervention area; 2 = Skaraborg County

======================

Introduction and Methods from paper by Svanstrom et al.:

I N T R O D U C T I O N

During the 1950s as many as 400 children were fatally injured annually in Sweden. This number has since fallen to include about 100 children 0-14 years who die from accidental injuries every year--roughly 40 girls and 60 boys.1 For every child killed there are about 100 children whose injuries are serious enough for them to receive inpatient hospital care. Among the fatalities in the preschool age group home and leisure injuries dominate, while most teenagers are killed in traffic.

In the middle of the 1970's the Safe Community concept was developed in Sweden and was first put into practice in Falks ping in 1975.2 Subsequently, in 1984, a similar programme was initiated in Lidkoping, Sweden. As with most other safe community programmes both began by establishing a local injury surveillance system. The purpose of this surveillance was to give information that would both help shape the intervention and assist in its evaluation.

The objectives of this paper are to describe the Lidkoping Accident Programme (LAPP); its possible effect on injury incidence; and to discuss how the processes might serve to reduce injuries over time.

The Lidkoping Accident Prevention Programme

In the early seventies, a community health unit was established to plan and coordinate health and safety promotion for Skaraborg County, including the Falks ping and Lidkoping municipalities. A fall of 34% in the incidence of injuries among preschool children was attributed to the Falks ping Accident Prevention Programme3-- from 48.6/1000 in 1978 to 32.2 in 1981/2. This inspired the local health authority in Lidkoping to start a similar safe community programme: LAPP.4 To raise the initiative's profile and to draw as much as possible on local knowledge and experience, an extensive intersectoral network was created.

The interventions agreed on including five elements: surveillance of injuries, provision of information, training, supervision, and environmental measures. The intervention started in 1984 and dealt with injuries affecting children and the elderly. This paper only addresses the former--injuries involving those under age 14 years.

M E T H O D S

STUDY AREAS

Skaraborg County, the home of Lidkoping, is located between Gottenburg and Stockholm in southern Sweden. It is mainly an agricultural and manufacturing county with 40% arable land compared with 8% for Sweden as a whole. In 1991 the population was 278,162. For this study comparisons are made between the intervention area, the municipality of Lidkoping (population 35,949), four bordering municipalities (population 42,078), and the whole of Skaraborg County. The 'border' municipalities use the same hospital as Lidkoping but received no intervention.

PROCESS EVALUATION

The LAPP evaluation involved studies of both process and outcome. The process evaluation was based mainly on reports made by the health planners, combined with newspaper clippings and interviews with key informants.

OUTCOME EVALUATION

The outcome evaluation was based on data from Skaraborg County Hospital discharge register. Cases are patients discharged from hospitals with an injury diagnosis coded E807-929 according to the International Classification of Diseases (ICD-9). These patients are then identified by place of residence regardless of the location of the hospital in which they were treated.

STATISTICAL METHODS

Difference in annual injury rates and their 95% confidence intervals between the comparison areas of Lidkoping, the four border municipalities combined, and Skaraborg County were calculated. Linear regression was then used to estimate the annual change in incidence.

=============

Exercise

I	Restrict your comparison to Girls in the intervention area versus Girls in the "4 Border municipalities" comparison area.
	a	Verify the beta estimates (-0.3 and 0.2). Do so (i) using the variable Year 'as is' as the x variable and (ii) using the variable Year-1987 as the x variable. Why is (ii) preferable?
	b	Obtain the SE of each beta estimate, and calculate a t-statistic to test equality of slopes.
	c	Fit master regression equations to the single dataset of 18 observations [intervention and comparison area] to represent (i) two parallel lines and (ii) two non-parallel lines. Report your conclusions. Note that the coefficient of the (area x year) interaction [i.e., product] term is the primary focus in this analysis.
II	Repeat steps a-c for Boys in the intervention versus "4 Border municipalities" comparison area.
III	Put the observations on boys and girls together in one dataset of 36 observations and repeat I c above. Why does it also help to include a term for gender? If one does not include a term for gender, why does the bigger dataset (36 observations) give a less statistically significant 'signal' than the gender-specific datasets with 18 observations?
IV	a In your analyses, you could model (1) the observed rates, with say Gaussian variations from the lines, taking no account of the sizes of the different denominators (2 ) the observed numbers of injuries, taking account of the sizes of the denominators, and using either i. Poisson or ii. extra-Poisson or iii. [constant] Gaussian variation from the fitted (expected) counts. b Comment, after trying each one, on the findings from these different analyses. Which one would you report as the primary finding, and why? [to reduce the number of analyses, compare these options in the single dataset that has 36 observations -- and a model that includes variation in rates by gender]. c [as a way to answer b] Read the letter from New Zealand (cf web site) and comment on their comments about the statistical analysis. Give your advice in a single page reply to a 'request for advice' from the Editor of this new journal [the report on the Lidkoping Accident Prevention Programme, with what the letter-writers suggest was an incorrect statistical analysis, appeared in the very first issue of the Journal]. This assignment is not is not contrived: the editor (Dr Barry Pless) did in fact consult JH on what he should do about the statistical issues raised by this at-first-glance-embarrassing letter.
V	Why do findings from ' additive rates' and 'multiplicative rates' models differ so little in this dataset? [again, fit the two in the single dataset of 36]

jh 2002.10.08